In this lecture, you will learn how to create your first Azure AI Foundry resource step by step. By the end of this session, you will not only understand how to set up Azure AI Foundry Services but also know how to configure essential components such as resource groups, subscriptions, and default project setups.

You will gain hands-on experience in:

• Navigating to Azure AI Foundry and creating a new resource.

• Configuring subscription types, resource groups, and deployment regions.

• Reviewing and validating configurations before deployment.

• Understanding the default resources and projects automatically created with Foundry.

• Preparing your environment for further exploration of Generative AI services and other Azure AI capabilities.

This lecture ensures you are ready with a properly deployed Azure AI Foundry environment so you can smoothly move forward with building, testing, and scaling AI-powered solutions in Azure.

If you are looking to get started with Azure AI Foundry, this lecture sets the foundation by walking you through the entire resource creation process in a clear and practical way.

In this lecture, you will explore the Azure AI Foundry Portal and learn how to navigate its key features step by step. By the end of this session, you will have a solid understanding of the core components inside Azure AI Foundry and how to use them to start building and managing your AI projects.

You will be able to:

• Navigate the Azure AI Foundry Portal and understand the default project setup.

• Locate and manage API Keys and Endpoints for Azure AI, OpenAI, and speech services.

• Explore the Model Catalog to find deployable models for generative AI and other AI services.

• Use the Playground to experiment with chat, summarization, and generative AI features.

• Understand how to create and manage AI Agents using built-in templates or custom setups.

• Learn the basics of fine-tuning models and integrating them into your projects.

• Monitor and evaluate generative AI applications through Observability and Optimization tools.

• Create and manage a Vector Store for RAG (Retrieval-Augmented Generation) applications.

• Access documentation and built-in Foundry Assistant for project support.

• Customize portal settings (themes, subscriptions, and account details) for your environment.

This lecture gives you a high-level overview of Azure AI Foundry, ensuring you are comfortable with its interface, tools, and project management features before diving deeper into model deployment and AI application development.

In this lecture, you will learn how to effectively use the Model Catalog in Azure AI Foundry to identify, compare, and select the right AI model for your business and technical needs. By the end of this session, you will understand how to explore different models, evaluate their performance, and prepare them for deployment in your AI applications.

You will be able to:

• Navigate the Model Catalog to discover available AI models for custom solutions.

• Use filters (collections, capabilities, inference tasks, and licenses) to quickly find models suited to your use case.

• Explore the Model Leaderboard to evaluate models based on quality, safety, cost, and throughput.

• Perform side-by-side model comparisons across metrics like latency, quality index, and cost.

• Understand the details of individual models, including capabilities, benchmarks, safety standards, and licensing.

• Identify top-performing models such as GPT-5, GPT-4, O-series, Mistral, and others to determine the best fit for your project.

• Prepare models for deployment as part of your Azure AI Foundry workflow (covered in upcoming lectures).

This lecture gives you the skills to make data-driven decisions when selecting AI models, ensuring that your applications are optimized for quality, safety, and cost efficiency within Azure AI Foundry.

In this lecture, you will learn how to deploy your first AI model in Azure AI Foundry. By the end of this session, you will be able to set up and configure a model deployment, access it through different programming languages, and start using it inside your applications.

You will be able to:

• Deploy a base model (GPT-5 Mini) in Azure AI Foundry.

• Configure deployment details such as model version, capacity, rate limits, and region.

• Validate and monitor your deployment process without errors.

• Access the deployed model using Python, JavaScript, REST APIs, Java, or C# with ready-to-use sample code.

• Retrieve deployment information including URI, authentication keys, and usage metrics.

• Test the deployed model directly inside the Azure AI Playground for chat and generative tasks.

This lecture equips you with the practical skills to deploy AI models in Azure AI Foundry, making them accessible for real-world applications and setting the foundation for advanced use cases in upcoming sessions.

In this lecture, you will learn how to explore and interact with your deployed AI model inside the Azure AI Foundry Playground. By the end of this session, you’ll clearly understand how to modify model instructions and context to generate more accurate, customized, and structured responses for different real-world use cases.

You’ll discover how simple changes in system prompts and parameters can completely transform the way your model behaves—whether you want it to act as a math tutor, a travel consultant, or a JSON data generator.

✅ What you’ll be able to do after completing this lecture:

• Navigate the Azure AI Foundry Playground effectively.

• Understand how model instructions influence responses.

• Modify system prompts to make the model act in different roles (e.g., tutor, assistant, consultant).

• Generate step-by-step solutions for problems like mathematical calculations.

• Create structured outputs such as JSON responses for technical tasks.

• Customize responses for use cases like travel itinerary planning, educational explanations, and more.

• Compare results by tweaking reasoning effort parameters and observe how response depth changes.

This lecture gives you hands-on experience in working with deployed AI models and prepares you to fine-tune system prompts for any professional or personal application.

? By the end, you’ll not only know how to interact with your deployed model effectively but also how to make it work smarter for your specific needs.

In this lecture, you will learn how to fine-tune key parameters of your deployed AI model inside the Azure AI Foundry Playground. By exploring these options step by step, you’ll understand how to control your model’s output length, reasoning ability, and summary generation to match your specific use cases.

We will experiment with three important parameters—Max Completion Tokens, Reasoning Effort, and Generate Summary—to see how they directly affect the responses produced by your GPT-5 Mini model. You’ll observe the difference between short vs. long outputs, fast vs. deeply reasoned answers, and responses with or without detailed summaries.

✅ What you’ll be able to do after completing this lecture:

• Adjust Max Completion Tokens to control response length and avoid truncated outputs.

• Configure Reasoning Effort (low, medium, high) to balance between speed and depth of reasoning.

• Enable and compare Generate Summary options (None vs. Detailed) for concise or in-depth outputs.

• Test real-world prompts like writing professional emails, solving math problems, and explaining scientific concepts to see parameter effects in action.

• Develop a strong understanding of how model parameters impact performance, accuracy, and response quality.

• Apply these techniques to optimize AI responses for business, education, or personal projects.

By the end of this lecture, you’ll have practical, hands-on experience in customizing AI model behavior using parameter tuning. This knowledge will help you generate more accurate, context-aware, and tailored outputs that fit your exact requirements.

In this lecture, you will learn how to configure and apply Safety System Messages in Azure AI Foundry Playground to ensure your deployed models generate responsible, compliant, and safe AI outputs. These safety features allow you to restrict harmful, ungrounded, or copyrighted content and prevent jailbreak manipulation, ensuring your applications remain secure and ethical.

We will explore how each safety option—Avoid Harmful Content, Avoid Ungrounded Content, Prevent Copyright Infringement, and Block Jailbreak Manipulation—adds additional system instructions to your model, and how these affect the responses you receive. You will also see how token usage increases when safety messages are applied and learn how to select only the safeguards you need.

✅ What you’ll be able to do after completing this lecture:

• Understand the role of Safety System Messages in controlling AI outputs.

• Configure options to filter harmful, racist, sexist, or violent content.

• Prevent models from generating fabricated or unverified (ungrounded) responses.

• Apply safety controls to block copyrighted material such as books, lyrics, or news articles.

• Implement safeguards against jailbreak and prompt manipulation attempts.

• Monitor how safety messages impact token usage and cost.

• Customize safety configurations based on your application’s requirements.

By the end of this lecture, you’ll be confident in applying safety system messages to your AI models so you can build applications that are trustworthy, ethical, and compliant while still delivering useful results.

In this lecture, you will learn how to use examples in prompt engineering to control the style, format, and structure of AI-generated responses. By the end of this session, you will clearly understand how to provide sample inputs and outputs so that the system generates answers in the exact way you want—whether it’s lists, structured formats, or customized text outputs.

What You’ll Learn in This Lecture:

• How to provide examples in prompts to guide AI behavior.

• Ways to format AI responses using structured examples (e.g., lists, categories, separators).

• How examples can improve accuracy, style, and consistency in generated responses.

• Practical demonstrations of transforming generic outputs into well-structured and customized answers.

• How to apply multiple examples for more refined and reliable results.

By completing this lecture, you will be able to design better prompts with examples that make your AI assistant more predictable, reliable, and aligned with your requirements. This is a crucial skill for anyone working in AI, ChatGPT, Azure AI Foundry, or advanced prompt engineering techniques.

In this lecture, you’ll learn how to use variables effectively inside your prompts to make them more dynamic, reusable, and flexible. By the end of this session, you will clearly understand how variables work, how to define them, and how to replace values within your prompts to create more personalized and adaptable outputs.

? Key Learning Outcomes:

• Understand what variables are and why they are important in prompt engineering.

• Learn how to create and assign values to variables within your prompts.

• Discover how to dynamically replace values (e.g., company names like Google or Microsoft) using variables.

• Apply variables in real-world scenarios to make your prompts more efficient and scalable.

• Gain confidence in structuring prompts for different use cases without rewriting the entire query.

This lecture is designed to help you save time, reduce redundancy, and improve flexibility while working with AI-powered prompts. Whether you are building chat assistants, automating workflows, or experimenting with AI text generation, mastering variables will give you a solid foundation for more advanced techniques.

In this lecture, you will learn how to deploy and work with Microsoft’s A54 reasoning models (such as A54-mini reasoning and chat completion models) in Azure AI Foundry. By the end of this session, you will not only understand how to deploy alternative models beyond ChatGPT, but also how to configure and control their behavior using parameters like max tokens, temperature, top-p sampling, and stop words.

You will also gain hands-on experience in:

• Deploying non-ChatGPT models from the Azure model catalog.

• Understanding the differences between GPT-based deployments and reasoning models.

• Using Python SDK and key-based authentication to integrate deployed models into your own applications.

• Running and testing models in the Azure Playground with real prompts.

• Deploying models as a web application with Entra ID authentication.

• Exploring ready-made prompts and assistants (like JSON formatter assistant) for structured responses.

• Managing resources by updating or deleting deployed models.

After completing this lecture, you will be able to confidently deploy, test, and integrate alternative AI models within Azure AI Foundry, giving you flexibility beyond ChatGPT for building real-world AI solutions.

? Perfect for learners preparing for AI-102: Microsoft Certified Azure AI Engineer Associate, or anyone looking to enhance their skills in AI model deployment, customization, and integration.

In this lecture, you’ll dive into the Azure AI Foundry Image Playground and explore how to work with powerful image generation and analysis models like DALL·E 3 and GPT-based multimodal models. By the end of this session, you will not only understand how to deploy and configure image-specific models but also how to interact with them through prompts and real-world use cases.

You’ll learn how to:

• Deploy and configure image models (e.g., DALL·E 3) within Azure AI Foundry.

• Generate AI-powered images from text prompts, experimenting with styles, quality, and vivid vs. natural outputs.

• Upload and analyze existing images by asking AI models contextual questions.

• Use multimodal capabilities in the Chat Playground to extract details, count objects, and generate creative descriptions from images.

• Manage deployments by creating, testing, and deleting models efficiently.

This lecture is designed to give you hands-on experience with generative AI for images, helping you understand how to use Azure AI Foundry’s Image Playground for both text-to-image generation and image-to-text analysis. After completing it, you will be confident in building interactive, creative, and analytical workflows using AI image models.

Whether you’re a beginner exploring generative AI or a developer building AI-powered applications, this lecture will equip you with practical skills to unlock the full potential of AI-driven image generation and interpretation.

In this lecture, you’ll discover how to create AI-generated videos using the Video Playground in Azure AI Foundry. We’ll explore how to deploy and use OpenAI’s Sora model—a powerful tool for video generation—similar to how DALL·E is used for images and ChatGPT for text.

By the end of this lecture, you will be able to:

• Navigate the Video Playground interface and understand its key features.

• Use pre-built prompts to generate diverse and creative AI videos.

• Customize video generation with aspect ratios, resolutions, and durations.

• Rewrite and enhance prompts with AI assistance for better video outputs.

• Generate high-quality, creative videos (e.g., fun examples like an elephant riding a bicycle while eating a burger).

• Manage and delete deployed models like Sora after experimentation.

This lecture gives you hands-on practice in AI-driven video creation while teaching you how to optimize prompts for the best results. You’ll gain the confidence to integrate AI video generation into your projects, whether for business, education, or creative storytelling.

? Perfect for learners exploring AI video tools, generative AI, and Azure AI Foundry Playgrounds.

In this lecture, you’ll explore the Audio Playground and learn how to transform text into high-quality audio using advanced AI-powered text-to-speech models. You will see step-by-step how to deploy an audio model, configure voice options, and generate speech in different formats like MP3, AAC, and Opus.

By the end of this lecture, you will be able to:

• Deploy and set up audio models in the playground.

• Generate realistic AI voices in various tones, styles, and speaking speeds.

• Apply custom model instructions (e.g., calm documentary narration, enthusiastic product pitch, or conversational tone).

• Choose between different voices and formats to best suit your project needs.

• Understand how to create engaging audio content for use in presentations, product demos, e-learning, podcasts, or creative projects.

This lecture gives you hands-on experience with the Audio Playground (Preview Mode) so you can leverage the power of text-to-speech (TTS) AI and bring your words to life with natural-sounding voices. Perfect for learners interested in AI tools, speech synthesis, and content creation.

In this lecture, you’ll explore the Microsoft Azure Speech Playground and learn how to leverage powerful speech-to-text, text-to-speech, and pronunciation assessment models for real-world applications. Unlike general AI models, Microsoft provides task-specific speech AI tools designed for accuracy and performance in professional scenarios.

By the end of this lecture, you will be able to:
✅ Use speech-to-text models for real-time transcription of audio recordings and uploaded files.
✅ Perform pronunciation assessments to identify mispronunciations, omissions, insertions, and speech clarity.
✅ Generate natural-sounding audio using text-to-speech models and explore multilingual voice galleries.
✅ Understand how to fine-tune and apply for professional and personal voice models for customized AI voices.
✅ Explore real-time video translation and integrate speech AI into your own applications using REST APIs, Python, or Java SDKs.

This lecture is designed for students preparing for Microsoft AI-102 certification as well as professionals interested in speech AI, voice technology, and real-time audio processing. You’ll gain practical, hands-on knowledge to implement speech capabilities in business applications without building models from scratch.

In this lecture, you will explore two powerful tools inside Microsoft’s Playground suite — the Language Playground and the Translator Playground. By the end of this session, you will gain hands-on experience with practical Natural Language Processing (NLP) tasks and understand how to leverage AI for text analysis, summarization, classification, and translation.

You’ll learn how to:

• Perform entity extraction to identify people, dates, products, and locations from text.

• Use summarization models to condense large text documents into clear, concise summaries.

• Apply sentiment analysis to detect positive, negative, or neutral tones in social media posts, reviews, and feedback.

• Detect and classify languages automatically with AI-powered models.

• Work with fine-tuned NLP models designed for specific use cases.

• Translate text and entire documents between multiple languages using the Translator Playground.

• Auto-detect source languages and convert them into your preferred target language with just a few clicks.

This lecture is perfect for anyone looking to apply AI-driven language processing in real-world applications, whether for business insights, research, content creation, or multilingual communication. After completing this lecture, you’ll be able to confidently use Microsoft’s Playground tools to analyze, summarize, classify, and translate text with ease.

In this lecture, you will learn the best practices of Prompt Engineering and how to transform vague, generic prompts into clear, structured, and powerful instructions that deliver accurate results from Large Language Models (LLMs). Through real-world examples, you’ll see how small improvements in prompts can drastically enhance the quality of responses, reduce hallucinations, and make AI outputs more reliable.

By the end of this lecture, you will be able to:

• Write specific and contextual prompts that guide AI to generate precise answers.

• Control the output format, structure, and length of AI responses.

• Assign roles and personas (e.g., teacher, trainer, tutor) to shape the AI’s behavior.

• Add constraints and conditions to make AI-generated content more useful and relevant.

• Create step-by-step instructions, comparisons, and examples within your prompts.

• Apply prompts for different scenarios like learning, productivity, fitness, exam preparation, or business use cases.

• Encourage reasoning and critical thinking in AI-generated responses.

This lecture is perfect for anyone who wants to master Prompt Engineering fundamentals, whether you are a student, developer, educator, or professional looking to leverage AI effectively in your work or learning.

In this lecture, we explore the newly introduced Grok family of models inside Azure AI Foundry. Grok, developed by Elon Musk’s company, is designed to handle complex reasoning tasks, larger context windows, and advanced use cases like document analysis, research, and code repository scanning.

You will learn:

• How to navigate the Azure AI Foundry model catalog and locate Grok models

• The deployment process for Grok-4 Fast Reasoning model in Azure

• How to test and interact with Grok inside the Azure Playground

• How to view and use the generated deployment code in Python applications

• Key advantages of Grok-4 compared to other models, including memory handling and large-scale text processing

By the end of this session, you will understand how to quickly deploy and start experimenting with Grok-4 in Azure, while also gaining insight into its unique strengths for real-world applications.

This lecture is perfect for developers, AI enthusiasts, and professionals who want to explore the capabilities of Grok inside Azure AI Foundry and integrate it into their projects.

In this lecture, you will learn how to create your very first AI Assistant using the Assistant Playground. We’ll walk step-by-step through deploying the right model (such as GPT-4), setting up assistant instructions, and customizing it to act as a fun, friendly tutor for kids. By the end of this session, you’ll clearly understand how to design an AI that follows your instructions, responds in a playful way, and stays focused on its assigned domain.

✅ What you’ll learn in this lecture:

• How to deploy and configure models in the Assistant Playground.

• The difference between using a standard model vs. a customized assistant.

• How to create a child-friendly tutor assistant that teaches subjects like math, science, language, and general knowledge in an engaging way.

• Writing effective assistant instructions to control tone, style, and scope of responses.

• Testing your assistant with prompts and comparing results with a generic model.

• How assistants handle domain restrictions, ensuring they only respond to relevant queries.

This lecture will give you the practical skills to build AI assistants tailored for specific use cases—whether for education, customer support, or creative tasks. After completing this lecture, you’ll be confident in setting up, testing, and refining AI assistants that align with your goals.

In this lecture, you will learn how to create a Smart Travel Buddy Assistant using GPT-4 models and enhance it with file search functionality for personalized travel planning. By the end of this session, you will be able to:

• Build a custom AI travel assistant that responds in a friendly, conversational tone, just like a professional travel guide.

• Integrate file search with vector stores to connect your assistant with external documents, such as travel guides or itineraries.

• Upload and vectorize knowledge bases (e.g., Paris Travel Guide) to provide accurate, context-based answers.

• Generate personalized travel itineraries (such as a 3-day trip plan for Paris) with the help of uploaded resources.

• Reference specific documents during chats, allowing your assistant to answer with precise details about destinations, attractions, food, culture, and travel tips.

• Test and compare generic vs. document-based responses to ensure your assistant uses the knowledge base effectively.

This lecture is perfect for learners who want to explore advanced features of GPT-4 assistants, including file search, embeddings, and real-world use cases in travel planning. By the end, you’ll have the skills to build an AI-powered guide that can plan trips, recommend attractions, and provide culturally rich insights—all backed by your own uploaded travel documents.

In this lecture, you will learn how to create a custom AI-powered Data Analysis Assistant (Data Analyzer Pro) using the Code Interpreter feature. By the end of this session, you will be able to confidently upload and analyze structured data files such as CSV, JSON, PDF, HTML, and code files to extract meaningful insights.

Through practical demonstrations, you’ll explore how to:

• Upload and analyze a student scores dataset in CSV format.

• Perform data cleaning, column summarization, and statistical computations automatically.

• Generate data visualizations such as heatmaps, box plots, and correlation charts.

• Identify patterns and relationships between different data columns.

• Retrieve insights like top-performing students based on average scores.

• Understand how the assistant limits itself to data and code-related tasks for accuracy.

By completing this lecture, you’ll gain hands-on skills in AI-driven data analysis and learn how to use a smart assistant to handle repetitive coding and analysis tasks. This will save you time, enhance your productivity, and prepare you for advanced projects involving data science, Python coding, and business intelligence.

Whether you are a beginner exploring AI tools or a professional looking to automate analysis workflows, this lecture will give you practical, real-world applications to leverage AI as your personal Data Analysis Buddy.

In this lecture, you will learn how to create and work with a Python Lab Assistant that can interpret, analyze, and execute Python scripts for real-world data science tasks. Building upon previous exercises with CSV files, this session introduces the process of uploading and running a Python file (customer_churn_demo.py) using the assistant.

By the end of this lecture, you will be able to:

• Upload and execute Python scripts within the assistant environment.

• Understand how the assistant analyzes Python code and explains functionality step by step.

• Automatically generate datasets, visualizations, and artifacts such as confusion matrices, feature importance plots, and exploratory data analysis (EDA) outputs.

• Download and review files generated from script execution (CSV, PNG, and text files).

• Query the assistant to display data samples (e.g., first 10 records) and statistical summaries from newly created files.

• Confidently use a code interpreter assistant to handle Python-related tasks, from running scripts to exploring generated outputs.

This lecture provides a hands-on, practical approach to Python code execution and data analysis using an AI-powered assistant, making it a valuable step in your journey to mastering data science automation.

In this lecture, you’ll learn how to add and use functions in the Assistant Playground to make your personal AI assistant more powerful and capable of handling real-time tasks. By the end of this session, you will be able to:

✅ Understand the importance of functions in GPT-based assistants and why they are required for fetching live data.
✅ Create and configure functions for different use cases such as:

• Weather Information Retrieval (Get Weather function)

• Stock Price Lookup (Get Stock Price function)

• Unit Conversion (Convert Unit function)
  ✅ Implement intent classification so that the assistant can decide which function to trigger based on the user’s query.
  ✅ Pass and manage parameters within functions (like location, units, or conversion values).
  ✅ Connect your assistant to external APIs and real-time data sources for practical, dynamic responses.

This lecture will help you bridge the gap between static responses and live, interactive AI-driven applications. After completing this lesson, you will be confident in designing assistants that can automatically trigger the right function to provide weather updates, financial data, or custom conversions — making your AI assistant smarter and more useful in real-world scenarios.

In this lecture, you will learn how to deploy a Retrieval-Augmented Generation (RAG) application in Azure AI Foundry step by step. We’ll begin by exploring the limitations of default models like GPT-5 Mini and then move on to deploying a RAG-enabled model that allows you to integrate your own custom data sources.

By the end of this lecture, you will be able to:

• Deploy a RAG-supported chat completion model in Azure AI Foundry.

• Create and configure an Azure Blob Storage account to upload and manage PDF documents.

• Enable CORS settings to securely allow AI models to access external data sources.

• Set up and configure Azure AI Search services for indexing and semantic search.

• Index multiple PDF documents (e.g., travel guides) and perform semantic + keyword-based queries.

• Build a complete RAG pipeline where your AI model retrieves relevant information from your own documents and provides references for responses.

• Compare responses between a base GPT model and a RAG-augmented model to see the benefits of retrieval-based augmentation.

• Clean up and manage resources such as storage accounts and AI services after deployment.

This lecture is designed for students and professionals who want to master Retrieval-Augmented Generation in Azure AI, enabling them to build enterprise-ready AI applications that combine private business data with powerful large language models.

? By completing this lecture, you will gain hands-on knowledge of deploying RAG applications in Azure, integrating custom data, and optimizing AI responses with Azure AI Foundry, Blob Storage, and AI Search Services.

In this lecture, you will learn how to fine-tune a base Large Language Model (LLM) in Azure AI Foundry using your own private dataset. Instead of relying only on general-purpose models like GPT-4, you’ll discover how to create a customized AI model tailored to your business needs—for example, customer support, order tracking, or invoice generation.

You will get hands-on experience with:

• Understanding different fine-tuning approaches (Supervised, Reinforcement, Direct Preference Optimization).

• Preparing and uploading training data in JSONL format.

• Fine-tuning a GPT-based model step by step inside Azure AI Foundry.

• Deploying your fine-tuned model to production.

• Testing real-world scenarios where your custom-trained model outperforms a default model.

• Managing and cleaning up deployments in Azure AI Foundry.

By the end of this lecture, you will be able to:
✔️ Fine-tune pre-trained models with your own data in Azure AI Foundry.
✔️ Deploy and test your fine-tuned model in a live environment.
✔️ Build AI solutions that deliver domain-specific, accurate, and contextual responses.
✔️ Optimize model performance for tasks like customer support, order management, and invoice handling.

This lecture is perfect for anyone looking to master fine-tuning techniques in Azure AI, enhance AI accuracy with private datasets, and gain practical deployment experience.

In this lecture, you will learn how to apply guardrails and content filters to your Generative AI applications in Azure AI Foundry. We will explore how default filters work on both text and image inputs, and how they help ensure safe, ethical, and compliant AI outputs.

You’ll see hands-on demonstrations using GPT-4 and GPT-5 models to understand how AI responds to sensitive prompts, including categories such as:

• Violence

• Self-harm

• Sexual content

• Hate speech

• Protected content (e.g., copyrighted material)

By the end of this lecture, you will be able to:
✅ Understand how default guardrails and content filters are applied in Azure AI Foundry.
✅ Analyze how AI models respond to both safe and unsafe prompts.
✅ Configure filter thresholds (Low, Medium, High) for different risk categories.
✅ Test text- and image-based inputs for policy compliance and safety.
✅ Identify how severity levels affect whether content is allowed or blocked.

This session gives you the foundation to implement responsible AI practices by applying built-in safeguards before moving on to creating your own custom filters and blocklists.

If you are working with Generative AI, Azure AI Foundry, or ChatGPT models, this lecture is essential to ensure your applications remain safe, ethical, and production-ready.

In this lecture, you will learn how to create and apply custom content filters in Azure AI Foundry to strengthen guardrails and controls for your generative AI applications. Building on the default filters explored earlier, we will go step-by-step through designing custom filters for both input and output content, configuring categories like violence, sexual content, and self-harm, and adjusting severity levels to match your application’s needs.

You will also discover how to set up and manage custom blocklists—allowing you to block specific words or phrases using exact matches or regex rules. By the end of this lecture, you will be able to:

• Configure custom input and output filters in Azure AI Foundry.

• Adjust severity levels to fine-tune content blocking for sensitive categories.

• Create and apply custom blocklists to filter unwanted words or terms.

• Test and validate your custom filters in the Azure AI playground with real-world prompts.

• Strengthen AI safety and compliance by applying advanced guardrails to GPT-4 and other models.

This hands-on knowledge will give you the confidence to enforce responsible AI practices and ensure your generative AI solutions meet compliance, safety, and ethical standards.

In this lecture, you’ll learn how to perform model evaluation in Azure AI Foundry using both manual and automated evaluation techniques. This step-by-step session will guide you through creating an AI Hub resource, deploying a model (such as GPT-4), and testing it with real datasets to measure performance and accuracy.

We start by setting up an AI Hub resource, which is required to enable the evaluation feature in Azure AI Foundry. You’ll then see how to prepare and upload a dataset in JSONL format, containing questions and expected responses. Using this dataset, we explore two approaches:

• Manual Evaluation: Compare the model’s generated responses with expected outputs and provide ratings (thumbs up or down) to measure response quality.

• Automated Evaluation: Use built-in evaluation metrics such as similarity scores, F1 score, and Likert scale ratings to automatically assess model accuracy, fluency, and relevance.

You will also learn how to configure evaluation criteria, run the evaluation process, analyze results, and export reports for further insights.

By the end of this lecture, you will be able to:

• Understand manual vs. automated model evaluation.

• Create and configure Hub resources in Azure AI Foundry.

• Deploy models (e.g., GPT-4) for evaluation.

• Prepare and manage test datasets for model performance checks.

• Analyze evaluation metrics and export results for reporting.

This lecture is designed to give you hands-on experience with evaluating generative AI models in Azure, helping you make informed, data-driven improvements to your projects.

In this lecture, you will learn how to integrate Azure OpenAI models into Python applications and build both command-line and GUI-based chat applications step by step. You will see how to set up a virtual environment, manage dependencies, configure environment variables, and authenticate with Azure to run AI-powered applications seamlessly.

By the end of this lecture, you will be able to:

• Create and activate a Python virtual environment for AI projects.

• Install and manage required dependencies using requirements.txt.

• Configure .env files with project endpoints, model deployments, and authentication details.

• Connect GPT-5 Mini for text generation and DALL·E 3 for image generation in Python apps.

• Run a command-line chat application using Azure OpenAI models.

• Build a GUI-based chat application with Gradio for interactive user experiences.

• Perform authentication with Azure using tenant ID and subscription for secure access.

• Test real-time Q&A scenarios with follow-up conversations powered by LLMs.

This lecture is ideal for anyone who wants to combine Python development with Azure AI Foundry, build real-world AI applications, and gain hands-on experience in deploying LLM-based chat systems.

In this lecture, you will learn how to transform a text-based Generative AI chat application into a GUI-powered application using the Gradio library. Building on the previous lesson where we executed the chatbot through the command line, this session demonstrates how to make your AI chatbot more interactive, user-friendly, and accessible via a web interface.

By the end of this lecture, you will be able to:

• Understand how to integrate Gradio to add a graphical user interface to your AI projects.

• Launch and interact with a Generative AI Chatbot through a browser (localhost with port access).

• Manage system messages, user prompts, and responses in a clear visual layout.

• Compare the command-line vs GUI versions of the same chatbot application.

• Build a full-fledged chat application that supports continuous conversation with follow-up questions.

This hands-on lecture equips you with practical knowledge to take your AI/ML projects beyond the terminal, making them accessible to both developers and non-technical users. Whether you’re aiming to build prototypes, demos, or production-ready apps, this GUI-based chatbot will give you the foundation to scale your AI solutions effectively.

In this lecture, you will learn how to extend a simple text-based chatbot into a multimedia chat application with image processing capabilities. You will see both a command-line version and a GUI-based version of the application, where you can upload or attach images and interact with the chatbot by asking questions about those images.

By the end of this lecture, you will be able to:

• Configure environment variables (.env file) for project endpoint and model deployment.

• Run a chatbot that analyzes images and provides descriptive responses.

• Work with both predefined image URLs and locally uploaded images.

• Ask complex queries about images (e.g., nutritional content of food items).

• Use the GUI version of the chat app to drag-and-drop images or capture snapshots via webcam.

• Understand how to integrate AI-powered image understanding into your Python applications.

This lecture is ideal for students who want to explore computer vision integration with AI chatbots and build interactive, real-world applications that go beyond text-based interaction.

In this lecture, you will learn how to integrate DALL·E 3 for AI image generation into your Python applications using Azure AI. We will start by deploying the DALL·E model in Azure AI Foundry and then connect it to our Python code with the required environment variables such as the model endpoint and API version.

You’ll see how to create AI-generated images from text prompts directly in Python—for example, generating fun images like a robot playing with a dog in a garden. We’ll also build a GUI-based image generator using Gradio, making the application more interactive and user-friendly.

By the end of this lecture, you will be able to:

• Deploy the DALL·E model in Azure AI.

• Configure environment variables for image generation.

• Write Python code to generate images using DALL·E 3.

• Create a Gradio-powered GUI for text-to-image generation.

This hands-on session will give you practical skills to integrate generative AI models into both code-based and graphical applications, helping you build your own AI image generation projects with Python.

In this lecture, you will learn how to build a Python-based AI Assistant Application with Azure AI Foundry. Unlike the earlier GUI-based assistant, this session takes you step by step through creating a fully functional assistant using GPT-4 deployment, configuring system instructions, and integrating it into your Python environment.

What you’ll learn in this lecture:

• How to create and deploy a GPT-4 model in Azure AI Foundry Assistant Playground

• How to set up environment variables (.env file) for API keys, endpoints, and API versions

• How to configure assistant instructions to act as a kids-friendly tutor for rhymes, quizzes, and fun learning activities

• How to reject unrelated queries with proper assistant instructions

• How to run and test your assistant application directly from Python code in VS Code

• How to use the auto-generated Azure Playground code in your own projects

By the end of this session, you’ll have a working AI Assistant in Python, powered by Azure AI Foundry and GPT-4, that can be customized for real-world use cases like education, tutoring, or interactive chat-based applications.

This lecture is perfect for:

• Developers who want to integrate Azure OpenAI GPT-4 into their Python projects

• Learners interested in AI assistant applications and automation

• Beginners exploring AI in Python with practical examples

In this lecture, we will create our first AI Agent using Azure AI Foundry — a Travel Reimbursement Agent capable of answering policy-related questions and processing travel claims automatically. You’ll learn how to combine knowledge files, Python code actions, and system instructions to build an intelligent, task-specific agent from scratch.

What You’ll Learn

• How to create and configure your first AI Agent in Azure AI Foundry

• How to deploy a base GPT model for your agent

• How to set system-level instructions and agent descriptions for real-world use cases

• How to attach knowledge documents (vector databases) like travel policies for contextual responses

• How to connect Python-based actions using the Code Interpreter for automation tasks

• How to test your agent inside the Playground and review performance metrics (input/output tokens, AI quality, completion time)

• How to make your agent handle corporate-level travel reimbursements using both document data and code execution

By the end of this lecture, you’ll have a fully functional Travel Reimbursement AI Agent that can read policy documents, process claims, and even generate reimbursement summaries using Python.

This session is perfect for:

• Developers learning how to build intelligent agents with Azure AI Foundry

• AI enthusiasts exploring multi-tool agent capabilities (knowledge + code interpreter)

• Professionals interested in automating corporate workflows using AI

In this lecture, you’ll learn how to create a Data Analysis AI Agent using Azure AI Foundry that can automatically analyze datasets, calculate key statistics, and generate insights from CSV files using Python.

You will build a Data Analysis Agent that leverages the Code Interpreter to read CSV data, perform statistical computations, and visualize results — all directly inside Azure AI Foundry. This hands-on session shows how to integrate data processing, reasoning, and chart generation into a single AI-powered workflow.

What You’ll Learn

• How to create a Data Analysis Agent in Azure AI Foundry

• How to use Code Interpreter to process CSV datasets

• How to perform data analysis using Python and Pandas within your AI agent

• How to compute averages, rankings, and correlations across multiple subjects or data columns

• How to generate charts and visual insights using Matplotlib

• How to explore and inspect run information, thread logs, and tool interactions for deeper debugging and understanding

• How to automate analytical tasks without manually writing Python scripts

By the end of this lecture, you’ll have a fully functional AI Data Analysis Agent capable of reading and analyzing real datasets, calculating metrics, and even visualizing results with minimal manual effort.

This session is perfect for:

• Developers and data analysts who want to combine AI and data analysis in Azure

• Learners looking to automate CSV data analysis using GPT models and Python

• Professionals exploring how AI Agents with Code Interpreter can simplify real-world data analytics workflows

In this lecture, you’ll learn how to create and interact with AI Agents programmatically using Python and Azure AI Foundry. Unlike the previous manual method from the Azure portal, this session focuses on building an agent completely through Python code — from creation to execution.

You’ll see how to automate the full lifecycle of an agent, attach files, configure system instructions, and interact with it directly in code. This hands-on approach helps you understand how to integrate Azure AI Agents into your real-world Python applications.

What You’ll Learn

• How to create an AI Agent using Python and Azure AI Foundry SDK

• How to configure your project endpoint, model deployment, and .env environment file

• How to attach data files (like .txt or .csv) as part of your agent configuration

• How to use the Code Interpreter to analyze attached data dynamically

• How to send and process prompts programmatically via Python

• How to generate insights like highest cost category or text-based bar charts using AI

• How to manage, test, and delete agents automatically from your code

By the end of this lecture, you’ll have a fully functional Python application that creates, configures, and interacts with Azure AI Agents — enabling automation, intelligent data analysis, and dynamic response generation right from your Python scripts.

This session is ideal for:

• Developers who want to automate agent creation using the Azure AI Foundry SDK

• Python programmers looking to integrate AI agents into their applications

• Learners exploring AI-driven data analysis and automation workflows

In this lecture, you’ll learn how to create a custom AI Agent in Azure AI Foundry using Python, complete with user-defined functions that automate real-world tasks. Unlike previous agents, this one includes a custom function (“Submit Support Ticket”) that acts as an intelligent action—allowing the agent to automatically handle support requests, generate tickets, and respond interactively.

You’ll build both CLI (command-line) and GUI (Gradio-based) versions of the application, giving you hands-on experience in integrating Azure AI Agent SDK with custom logic inside Python.

What You’ll Learn

• How to create a new AI Agent via Python code using the Azure AI Foundry SDK

• How to define and attach a custom function as an agent action

• How to build a “Technical Support AI Agent” that collects user issues and submits tickets

• How to manage agent creation, testing, and deletion through code

• How to build a GUI version using the Gradio library for an interactive user experience

• How to configure environment variables (.env file), endpoints, and model deployments

• How to handle authentication and integration with Azure AI Foundry seamlessly

By the end of this lecture, you’ll have a working Custom Function AI Agent that can execute real tasks — like creating and storing support tickets — using both Python scripts and a graphical interface.

This session is perfect for:

• Developers who want to automate workflows with AI Agents in Azure

• Python programmers exploring custom tool integration in AI models

• Professionals building AI-powered customer support systems or other task-specific assistants

In this lecture, you’ll learn how to build a Multi-Agent AI System using Azure AI Foundry and Python — a system where multiple specialized agents work together to deliver intelligent, coordinated results. Unlike previous single-agent setups, this session demonstrates how one main (orchestrator) agent can delegate tasks to connected sub-agents, combine their responses, and produce a unified, intelligent output.

We’ll design a fun and practical project — the “Movie Night AI Assistant” — where different agents collaborate to recommend movies, snacks, and fun facts for a perfect movie night. You’ll see both CLI-based and Gradio GUI-based versions of this multi-agent system in action.

What You’ll Learn

• How to create a multi-agent architecture in Azure AI Foundry using Python

• How to connect and coordinate multiple AI agents (orchestrator + sub-agents)

• How to define specialized roles for each agent (e.g., genre selection, snack suggestion, fun facts)

• How the root agent orchestrates sub-agent communication and combines their outputs

• How to create an interactive GUI version with Gradio for user-friendly execution

• How to manage agent creation, deletion, and runtime behavior programmatically

• How to enhance AI projects with real-world orchestration workflows

By the end of this lecture, you’ll have a complete understanding of how to build, connect, and manage multi-agent systems in Azure AI Foundry. You’ll also learn how to turn these agents into an engaging, user-ready application using Python and Gradio.

This session is perfect for:

• Developers who want to design collaborative AI agent systems

• AI engineers exploring multi-agent orchestration in Azure AI Foundry

• Learners building real-world AI assistants using GPT-based models and Python

In this lecture, you’ll learn how to set up and work with Prompt Flow in Azure AI Foundry, enabling you to design and test conversational AI experiences with greater flexibility. Since the default project type does not support Prompt Flow, you’ll discover how to create a new AI Hub project, configure resources, and assign the right permissions for seamless integration.

By the end of this lecture, you will be able to:

• Create and configure an AI Hub project to enable Prompt Flow in Azure AI Foundry.

• Assign and manage system-assigned identities and roles to ensure proper access to storage accounts.

• Build and customize chat flows and configure inputs like chat history and user queries.

• Deploy and test a chatbot application powered by models such as GPT-4.1.

• Troubleshoot common deployment issues and manage compute resources efficiently.

This lecture is ideal for students who want hands-on knowledge of building, testing, and deploying chat-based AI applications using Prompt Flow in Azure AI Foundry. After completing this session, you’ll have the skills to design interactive AI workflows and deploy them as endpoints for real-world applications.

In this lecture, you will learn how to create your very first Prompt Flow in Azure AI Foundry—a powerful no-code way to design, test, and manage AI workflows. Without writing any code, you’ll explore how to build a basic end-to-end flow that connects inputs, outputs, and language model prompts in a structured pipeline.

By following along, you will:

• Understand the concept of Prompt Flows and why they are essential for building AI-powered applications.

• Learn how to set up inputs and outputs for your flow.

• Configure a serverless compute session to power your workflow.

• Create and deploy a language model (GPT) to generate meaningful responses.

• Design a simple flow that takes a topic (e.g., smartphone) and returns five interesting facts about it—scalable to any subject you choose.

• Gain hands-on experience with Azure AI Foundry’s UI tools for managing and validating your flows.

By the end of this lecture, you’ll be confident in building your first AI workflow using Prompt Flows, setting the foundation for more advanced projects like chatbots, knowledge assistants, and intelligent business solutions.

This session is perfect for beginners who want to get started with Azure AI Foundry, explore prompt engineering, and leverage low-code/no-code tools for AI development.

In this lecture, you will learn how to create, validate, and run a simple flow in Azure AI Studio using compute sessions. We will walk through step-by-step how to configure input parameters, define LLM blocks, handle UI issues, and validate your inputs to ensure smooth execution of your prompt flow.

By the end of this lecture, you will be able to:

• Start and manage compute sessions in Azure AI Studio.

• Define and validate input parameters for prompt flows.

• Configure LLM blocks with custom settings such as max tokens, response type, and API selection.

• Debug and troubleshoot common UI issues that may occur while building flows.

• Run and test your first complete prompt flow, view outputs, and analyze execution logs.

• Understand how to deploy your flow for real-world applications with Python integration or direct endpoint usage.

This lecture is designed to give you hands-on experience with building and executing flows in Azure AI Studio, preparing you to confidently deploy and integrate flows into your AI-driven applications.

In this lecture, you will learn how to successfully test, monitor, and manage your first deployed AI flow within Azure AI Studio. We’ll walk through the complete process of verifying deployment status, understanding key parameters such as provisioning state, Swagger URI, target URI, and REST endpoints, and using the provided primary key for secure testing.

You will also explore how to test your deployment directly from the JSON editor or integrate it seamlessly into applications using Python, JavaScript, C#, or JSON code snippets. By the end of this session, you’ll not only validate your deployment but also understand how to monitor its performance, manage costs, and clean up resources by deleting unused endpoints.

✅ What you’ll be able to do after completing this lecture:

• Verify and monitor the status of your AI model deployment.

• Use Swagger and REST endpoints to test deployed models.

• Integrate your endpoint into applications using Python, JavaScript, or C#.

• Test input variables and view real-time results inside Azure AI Studio.

• Manage deployment costs and clean up unused endpoints.

This lecture ensures you gain hands-on experience with real-world AI deployment workflows, preparing you to confidently create, monitor, and optimize your own Azure AI solutions.

In this lecture, you will learn how to design and implement a multi-step prompt flow using two Large Language Models (LLMs) in sequence. Instead of relying on a single LLM, you will explore how to chain models together so that the output of one becomes the input for the next—unlocking more powerful and flexible AI workflows.

By the end of this lecture, you will be able to:

• Clone and extend an existing basic prompt flow into a sequential flow with multiple LLMs.

• Configure LLM-to-LLM communication where the first model generates a structured explanation and the second model simplifies it for easier understanding.

• Set up inputs, outputs, and system prompts that allow your flow to adapt to different use cases (e.g., generating expert content vs. simplifying it for beginners or children).

• Deploy and run your sequence flow on Azure AI with optimized compute settings.

• Understand how sequential LLMs can transform complex technical explanations into simplified insights that anyone—even a 7-year-old—can understand.

This hands-on lecture will give you practical experience in building advanced prompt flows with multiple LLMs in Azure, preparing you to design intelligent workflows for real-world AI applications.

In this lecture, you will learn how to design and implement a parallel flow using multiple Large Language Models (LLMs) to generate and consolidate insights for real-world use cases like SWOT (Strengths, Weaknesses, Opportunities, Threats) analysis.

We will walk through creating a parallel workflow where:

• A single input (e.g., company name such as Microsoft or Google) is processed by four different LLMs simultaneously.

• Each LLM independently generates a specific part of the analysis (strengths, weaknesses, opportunities, threats).

• A final LLM combines these outputs into a concise SWOT summary.

• You’ll also learn how to capture and view intermediate outputs for validation and debugging.

By the end of this lecture, you will be able to:

• Build parallel AI flows in Azure Prompt Flow (or similar platforms).

• Connect multiple LLMs to handle different aspects of a single task.

• Generate structured insights like business analysis, market research, and summaries using AI pipelines.

• Combine multiple outputs into a final, polished result.

• Optimize compute sessions and manage flow execution efficiently.

This lecture is especially useful for learners who want to:

• Automate business intelligence tasks with AI.

• Understand parallel execution in AI workflows.

• Explore multi-model collaboration to improve results.

By mastering this approach, you’ll unlock the ability to create more scalable, accurate, and insightful AI-powered applications.

In this lecture, you will learn how to create and implement a Chat Flow in Azure AI Prompt Flow and understand how it differs from a standard flow. Unlike traditional flows that only process single inputs, Chat Flows maintain conversation history and context, allowing you to build intelligent chatbot-like applications powered by large language models such as GPT-4.

By the end of this lecture, you will be able to:

• Understand the difference between Standard Flow vs. Chat Flow in Azure AI.

• Configure a Chat Flow to retain and use past conversation history for more accurate responses.

• Connect Chat Flow with Azure AI Hub and GPT models for contextual Q&A.

• Experiment with scenarios where contextual memory impacts chatbot answers versus cases where no history is provided.

• Build the foundation for context-aware conversational AI applications within Azure AI Foundry.

This hands-on session is designed to give you practical knowledge of creating interactive AI-driven chat experiences, helping you move beyond simple prompt-response flows into building real-world conversational AI systems.

In this lecture, you will learn how to integrate Prompt Flow with Index Lookup to build a Retrieval-Augmented Generation (RAG) application using your own documents. Instead of relying solely on a Large Language Model (LLM), you will understand how to use vector indexing and document embeddings to fetch the most relevant information from personal or organizational files.

By the end of this lecture, you will be able to:

• Understand the concept of RAG (Retrieval-Augmented Generation) and why it’s essential for answering domain-specific queries.

• Create and manage vector indexes of documents such as PDFs to improve information retrieval.

• Deploy a search service in Azure for indexing and querying documents efficiently.

• Use Index Lookup to connect questions with relevant documents instead of overloading the LLM with unnecessary data.

• Build both Standard Flow and Chat Flow for document-based Q&A systems inside Azure Prompt Flow.

• Verify and test your indexed data to ensure accurate retrieval of answers from your knowledge base.

This lecture is perfect for learners who want to explore AI-powered search systems, enterprise document Q&A solutions, and real-world LLM applications using Azure OpenAI.

In this lecture, you will learn how to integrate Index Lookup with Azure AI Search into your Prompt Flow to build a more powerful and accurate Retrieval-Augmented Generation (RAG) application. Moving beyond a simple LLM-based flow, you will explore how to connect and configure Azure OpenAI embeddings, metadata fields, and semantic search configurations to retrieve the most relevant information from indexed documents.

By the end of this lecture, you will be able to:

• Configure Azure AI Search Indexes and set up embedding fields, metadata, and semantic configurations.

• Implement Index Lookup tools to fetch top results using hybrid (vector + keyword) queries.

• Connect retrieved outputs to an LLM (such as GPT-4) for context-aware and reliable responses.

• Build a complete RAG pipeline where user queries are answered using indexed document knowledge.

• Validate and test queries like “Why is hydration important for the body?” or “What are macronutrients and why are they important?” and analyze how results are retrieved.

This lecture equips you with the knowledge to design scalable AI solutions using index-based search and LLMs, setting the foundation for advanced chat-based flows with retrieval capabilities. Perfect for learners aiming to master Azure AI, OpenAI GPT models, and real-world RAG applications.

In this lecture, you will learn how to build a chat-enabled index flow that combines chat history with document indexing to deliver accurate and context-aware answers. Unlike a standard flow, this lecture focuses on creating an Index-Chat Flow, where the entire chat history and user queries are processed together with indexed data to generate meaningful responses through a chat interface.

By the end of this lecture, you will be able to:

• Create and configure a Chat Flow integrated with index lookup.

• Understand how to pass chat history and user questions into the flow for better context handling.

• Use Azure OpenAI embeddings (text-embedding-ada-002) for semantic search.

• Configure hybrid query types (vector + keyword search) for more accurate retrieval results.

• Build a retrieval-augmented generation (RAG) application that uses your indexed documents inside a chat interface.

• Deploy and validate your flow, ensuring responses are sourced from your documents instead of just the model’s general knowledge.

This hands-on session equips you with the skills to design intelligent conversational AI applications that can extract knowledge from your indexed data while maintaining natural dialogue capabilities.

In this lecture, you will learn how to integrate Google Search API (SerpAPI) into your Prompt Flow to fetch real-time, up-to-date information and enhance the accuracy of your AI applications. Unlike static LLM responses, this integration ensures that your prompt flows can retrieve the latest data directly from Google Search and pass it back to your language model for precise answers.

By the end of this lecture, you will be able to:

• Create a Prompt Flow with SerpAPI integration.

• Configure API keys, connections, and compute sessions within the flow.

• Debug and resolve common flow errors such as incorrect variable mapping.

• Fetch dynamic, real-time answers (e.g., population data, latest facts) using Google Search results.

• Combine Google Search output with GPT models to build AI solutions that are both intelligent and up-to-date.

This lecture is ideal for anyone who wants to go beyond static AI responses and build RAG (Retrieval-Augmented Generation) applications using Google Search API inside AI Foundry prompt flows. By completing this lecture, you will gain practical skills to bridge LLMs with real-time internet data—a must-have for developers, data scientists, and AI enthusiasts.

In this foundational lecture, you will gain a comprehensive overview of the first section of the course: Getting Started with Azure AI Services. This lecture sets the stage for the exciting journey into the world of Azure AI, highlighting key aspects you'll explore, including service creation, API interactions, and monitoring.

Key Points Covered:

• Introduction to Azure AI Services:

  • Overview of the Azure AI services and how they form the entry point for interacting with Azure AI components and APIs.

  • Importance of foundational services for enabling various AI functionalities.

• GitHub Repository Utilization:

  • Introduction to a publicly available GitHub repository created by Microsoft Learning.

  • Details on using Python SDK and C# SDK to interact with Azure AI services, with Python as the primary language throughout the course.

• Service Creation on Azure Portal:

  • Step-by-step guide to creating Azure AI services in the Azure portal.

  • Understanding the significance of these services for API interaction.

• API Interaction with Python Code:

  • How to interact with Azure AI APIs using Python.

  • Demonstrating ways to secure API interactions using Azure Key Vault for enhanced security.

• Monitoring Azure AI Services:

  • Techniques to monitor Azure AI services for specific parameters, such as total requests for a particular service like Computer Vision.

  • Setting up notifications for threshold breaches.

• Deployment and Safety Features:

  • Deploying Azure AI services as container applications.

  • Exploring Azure AI’s built-in safety features for detecting harmful, violent, or copyrighted content.

• Exciting Applications:

  • Practical examples and use cases to understand the capabilities and potential of Azure AI.

This lecture provides a theoretical foundation to help you understand key concepts and terminologies in artificial intelligence and their implementation using Azure services. Before diving into hands-on activities, we’ll explore the fundamentals of AI, its types, ethical principles, and Azure AI services. By the end of this lecture, you’ll have a solid understanding of AI basics, responsible AI practices, and the capabilities Azure offers to build intelligent solutions.

Key Points Covered:

1. Understanding Artificial Intelligence (AI):

• Definition: Simulation of human intelligence in machines for learning, reasoning, and problem-solving.

• Types of AI:

  • Narrow AI: Performs specific tasks (e.g., chatbots, sentiment analysis).

  • General AI: Aims to perform intellectual tasks across domains (futuristic).

  • Super AI: Hypothetical AI surpassing human intelligence.

2. Key AI Applications:

• Natural Language Processing (NLP)

• Computer Vision

• Speech Recognition

• Machine Learning

3. AI Terminologies:

• Machine Learning: Subfield of AI focused on pattern recognition and data-driven learning.

• Deep Learning: Uses artificial neural networks to mimic the brain’s structure for learning.

• Related terms: Model training, inference, NLP, and computer vision.

4. Responsible AI Practices:

• Principles of Responsible AI:

  • Fairness, reliability, privacy, inclusiveness, transparency, and accountability.

• Ethical considerations:

  • Monitoring AI’s societal impact.

  • Fairness audits and bias detection.

5. Azure Machine Learning Capabilities:

• End-to-end machine learning model creation, deployment, and monitoring.

• Key Features:

  • AutoML for automated model training.

  • Predictive analytics, fraud detection, and forecasting use cases.

6. Azure AI Services Overview:

• Prebuilt APIs for vision, speech, language, and decision-making tasks.

• Use cases:

  • Virtual agents, text analysis, image processing.

• Integration with OpenAI models for text generation and summarization.

7. Azure AI Search (formerly Cognitive Search):

• AI-powered search capabilities with features like semantic search and OCR integration.

• Ideal for knowledge mining and enterprise-level search applications.

In this practical lecture, we kick off our hands-on journey into Azure AI services. You will learn how to set up and provision your first Azure AI service using Visual Studio Code and a Microsoft-provided GitHub repository. By the end of this lecture, you will understand how to clone the repository, create Azure AI services, and interact with the APIs using REST and SDK interfaces.

Key Points Covered:

1. Cloning the Repository:

• Microsoft's GitHub Repository: Introduction to the public repository (MsLearn AI Services) containing pre-built lab files.

• Cloning Process:

  • Creating a local folder (e.g., EasyCog).

  • Cloning the repository using a command-line interface.

  • Importing the repository into Visual Studio Code.

2. Setting Up Visual Studio Code:

• Installation Guidance: Steps to download and install Visual Studio Code based on your operating system (Windows, Linux, or MacOS).

• Importing the Repository: How to open the cloned repository in Visual Studio Code for use throughout the course.

3. Provisioning Azure AI Services:

• Azure Portal Overview:

  • Navigating the Azure portal to access Azure AI services.

  • Creating a new resource group to organize cognitive services.

• Service Configuration:

  • Selecting the region and setting a unique name for the service.

  • Choosing the default pricing tier (Standard S0).

• Deployment Steps: Initiating and completing the deployment of Azure AI services.

• Accessing Keys and Endpoints:

  • Viewing and securing keys and endpoints for API usage.

  • Regenerating keys in case of compromise.

4. API Interaction:

• REST Interface and SDK:

  • Using the keys and endpoints to interact with various Azure AI APIs (e.g., Computer Vision, Language Translator, Content Safety).

  • REST and Python SDK as primary interaction methods.

5. Resource Cleanup:

• Cleaning up resources after completing the exercise to avoid unnecessary costs.

Objectives for the Lecture:

• Clone and set up the Microsoft-provided GitHub repository in Visual Studio Code.

• Provision and deploy Azure AI services on the Azure portal.

• Understand how to access and use API keys and endpoints for interacting with Azure AI services.

• Prepare for API usage via REST and SDK in upcoming exercises.

In this hands-on lecture, we demonstrate how to utilize the Azure AI services provisioned in the previous session within Python code. You'll learn to interact with these services using both the REST interface and the Python SDK. This lecture covers essential setup steps, including creating a virtual environment, installing required libraries, and configuring .env files for secure API key usage. By the end, you'll have successfully run language detection tasks using Azure AI services.

Key Points Covered:

1. Preparing the Environment:

• Python Installation:

  • Ensure Python (version 3.9 or later) is installed on your local machine.

  • Download and install Python for Windows, Linux, or Mac as required.

• Creating a Virtual Environment:

  • Steps to create a virtual environment (venv) in the cloned repository directory.

  • Activating the virtual environment for an isolated Python setup.

2. Configuring the Environment:

• Setting Up .env Files:

  • Store Azure AI service keys and endpoints securely in .env files.

  • Configure for both REST and SDK clients to ensure smooth API interactions.

3. Installing Required Libraries:

• Python Libraries:

  • Install python-dotenv for managing environment variables.

  • Install the Azure SDK library azure-ai-textanalytics (version 5.3.0) for SDK interactions.

4. Using the REST Client:

• Steps to Execute:

  • Run the rest_client.py script from the cloned repository.

  • Pass text inputs for language detection via REST API calls.

  • Understand how the API keys and endpoints are used in the requests.

• Examples:

  • Detect languages like English, French, Persian, and Japanese using the REST client.

5. Using the SDK Client:

• Steps to Execute:

  • Run the sdk_client.py script using the Azure SDK library.

  • Leverage ready-made functions like detect_language() to simplify API usage.

• Examples:

  • Detect languages using the SDK, comparing functionality with the REST client.

6. Key Differences Between REST and SDK:

• REST Client: Offers granular control over API requests and responses.

• SDK Client: Provides abstraction with prebuilt functions for easier interaction.

7. Cleaning Up Resources:

• Importance of deleting resources to avoid unnecessary costs.

• Steps to delete Azure AI services from the Azure portal (if not required further).

Objective for the Lecture:

• Set up a Python environment with all necessary dependencies for Azure AI service integration.

• Configure API keys and endpoints securely for REST and SDK client usage.

• Run language detection tasks using both REST API and SDK methods.

In this lecture, we focus on securing Azure AI services by managing API keys and enhancing authentication methods. You’ll learn to secure your API interactions, regenerate compromised keys, and understand the limitations of directly embedding keys in requests. The lecture concludes with an introduction to Azure Key Vault for advanced security practices.

Key Points Covered:

1. Introduction to API Key Security:

• Why Secure API Keys?

  • API keys embedded in requests can be intercepted, leading to security vulnerabilities.

  • Regenerating keys helps prevent misuse if they are compromised.

• Security Practices for API Key Management:

  • Avoid directly embedding keys in code or requests.

  • Use secure storage methods like Azure Key Vault.

2. Securing Azure AI Services:

• Testing with API Keys:

  • Using the Rest_test.cmd file to test Azure AI services via cURL.

  • Replacing placeholders with actual API keys and endpoints in .cmd files.

  • Running tests to detect languages using the API.

• Regenerating Keys:

  • Steps to regenerate keys from the Azure portal.

  • Validating that the old key is no longer functional after regeneration.

  • Updating API keys in the .cmd file to restore access.

3. Limitations of Direct Key Embedding:

• Key Exposure Risk:

  • Keys passed directly in requests can be exposed over the network.

  • This method poses a significant security risk for sensitive operations.

• Need for Enhanced Security:

  • Transition to Azure Key Vault for secure key management.

4. Introducing Azure Key Vault:

• Overview:

  • Azure Key Vault is a secure service for storing and managing sensitive data like API keys, certificates, and secrets.

• Next Steps:

  • Use Azure Key Vault to securely retrieve and manage keys in real-time.

  • Integration of Key Vault with Azure AI services to eliminate key exposure risks.

Objectives for the Lecture:

• Understand the importance of securing API keys for Azure AI services.

• Learn to test API functionality with keys and regenerate them when compromised.

• Recognize the limitations of embedding keys in requests.

• Prepare to use Azure Key Vault for advanced key management in the next session.

In this lecture, we dive deeper into securing Azure AI services by integrating Azure Key Vault. Learn how to securely store API keys as secrets in Key Vault, create a service principal for authentication, assign appropriate roles, and configure permissions. By the end of this session, you will understand how to access Azure AI services securely from your applications without exposing sensitive keys.

Key Points Covered:

1. Setting Up Azure Key Vault:

• Creating a Key Vault:

  • Steps to create a new Azure Key Vault in the Azure portal.

  • Configuring Key Vault settings such as region, resource group, and permissions model.

• Storing Secrets:

  • Adding API keys as secrets in the Key Vault.

  • Ensuring the secure storage of sensitive data.

2. Creating a Service Principal:

• App Registration:

  • Steps to create a service principal (app registration) in Azure Active Directory.

  • Noting critical details: Application ID, Object ID, and Tenant ID.

• Generating Client Secrets:

  • Creating a client secret for the service principal.

  • Storing the secret securely for use in the application.

3. Assigning Roles and Permissions:

• IAM Role Assignment:

  • Assigning the Contributor role to the service principal for accessing Key Vault.

  • Exploring alternatives for more granular access (e.g., Key Vault-specific roles).

• Key Vault Access Policy:

  • Configuring access policies for the service principal to get and list secrets.

  • Ensuring the application has the necessary permissions without over-provisioning.

4. Integrating Azure Key Vault with Applications:

• Python or Other Applications:

  • Utilizing Azure Key Vault SDK or REST API to retrieve secrets securely.

  • Ensuring sensitive data is no longer hardcoded in the application.

• Advantages:

  • Mitigating risks of key exposure during network transmissions or in the codebase.

5. Security Best Practices:

• Use service principals for authentication instead of embedding credentials in the application.

• Regularly regenerate and rotate keys for enhanced security.

• Monitor and audit Key Vault access logs to detect anomalies.

In this lecture, we implement Azure Key Vault integration with Python to securely access Azure AI services without exposing sensitive information such as API keys. This session demonstrates configuring a Python script to use Azure Key Vault, installing required libraries, and running a sample language detection API call securely. By the end of this lecture, you will understand how to leverage Azure Key Vault and service principals in Python for enhanced security.

Key Points Covered:

1. Setting Up Environment for Key Vault Integration:

• Configuration in .env File:

  • Add essential details like:

    • AI Service Endpoint (e.g., Language API endpoint).

    • Key Vault Name (e.g., myKV255).

    • Tenant ID, Application (Client) ID, and Application Password.

  • Ensure proper formatting without spaces.

2. Installing Required Python Libraries:

• Libraries required for Key Vault integration and API access:

  • Azure Key Vault Secrets (azure-keyvault-secrets).

  • Azure Identity (azure-identity).

  • Azure Text Analytics (azure-ai-textanalytics).

3. Running the Key Vault Integration Script:

• Key Vault Client Script:

  • A Python script (key_vault_client.py) configured to:

    • Fetch secrets (keys) securely from Azure Key Vault.

    • Use the retrieved secrets to call Azure AI services.

• Executing the Script:

  • Activate the Python virtual environment and run the script.

  • Example output: Securely detects the language of the input text without exposing keys directly in the code.

4. Key Benefits of Using Azure Key Vault:

• Eliminates the need to hardcode secrets in the application.

• Enhances security by leveraging Azure-managed secret storage.

• Provides centralized secret management and auditing capabilities.

5. Clean Up Resources:

• Best Practices for Resource Management:

  • Delete Azure Key Vault and other resources (e.g., app registrations) when no longer needed to avoid unnecessary costs.

  • Ensure the removal of access permissions to maintain a secure environment.

This lecture demonstrates how to monitor Azure AI services using alerts and metrics to ensure their effective operation and performance. Learn to configure alert rules, visualize metrics, and set up notifications for proactive monitoring. By the end of this session, you’ll understand how to create, configure, and visualize alerts and metrics to monitor Azure AI services efficiently.

Key Points Covered:

1. Configuring Alerts in Azure Portal:

• Accessing Monitoring Tools:

  • Navigate to the Azure AI service resource (e.g., Azure AI Start).

  • Open the "Alerts" section under monitoring tools.

• Creating an Alert Rule:

  • Select signals to monitor (e.g., total number of calls, errors, latency).

  • Set threshold values and evaluation periods for triggering alerts.

  • Example: Trigger an alert when total calls exceed 5 within a 5-minute period.

2. Setting Up Action Groups:

• Defining Actions:

  • Create an action group to handle alerts (e.g., sending email notifications).

  • Specify notification types, such as email, SMS, or triggering Azure Functions.

• Example Configuration:

  • Configure an action group with email notifications to the administrator.

  • Link the action group to the alert rule.

3. Testing Alerts:

• Sending API Requests:

  • Use a Python SDK client to send multiple API requests and exceed the alert threshold.

  • Example: Detect language using Azure AI Language APIs.

• Receiving Notifications:

  • Verify email notifications for triggered alerts.

  • Example: "Azure Activated Severity Alert" email indicates the alert was triggered due to exceeding the threshold.

4. Visualizing Metrics:

• Viewing Metrics in Azure Portal:

  • Explore predefined metrics such as total calls, client errors, and latency.

  • Use visualizations like line charts, bar charts, and area charts for better insights.

• Creating Custom Dashboards:

  • Save visualized metrics to a dashboard for continuous monitoring.

  • Example: Monitor average latency or call count trends over time.

5. Additional Use Cases:

• Custom Metrics and Alerts:

  • Monitor specific metrics such as client errors, rate limits, or image moderation calls.

  • Configure alerts for unusual activity or performance degradation.

• Continuous Monitoring:

  • Leverage dashboards for a holistic view of Azure AI service performance.

In this lecture, we explore how to deploy Azure AI services within containers for scalable, portable, and secure AI solutions. Containers enable efficient deployment of AI models across different environments, offering offline capabilities and enhanced performance. This session introduces the concept of containers, their benefits, and the workflow for deploying pre-built Azure AI service models using Docker and Kubernetes.

Key Points Covered:

1. Introduction to Containers:

• What are Containers?

  • Lightweight alternatives to virtual machines.

  • Self-contained packages with code, runtime libraries, and dependencies.

  • Portable across development, testing, and production environments.

• Why Use Containers for AI?

  • Simplifies deployment and scaling of AI models.

  • Enhances performance and debugging.

  • Maintains consistency across environments.

2. Key Concepts and Terminologies:

• Docker:

  • Platform for creating and managing container images.

• Container Images:

  • Templates used to create containers.

• Container Registry:

  • Centralized repository for storing and managing container images (e.g., Azure Container Registry).

• Orchestration Tools:

  • Tools like Kubernetes for managing container clusters.

3. Workflow for Container-Based AI Deployment:

1. Create Container Images:

   • Build the container image with all necessary dependencies.

2. Push to Container Registry:

   • Upload the image to Azure Container Registry.

3. Deploy Containers:

   • Deploy locally or remotely using Docker or Kubernetes.

4. Manage Containers:

   • Monitor, scale, and update containers based on workload or application requirements.

4. Azure AI Services and Containers:

• Pre-Built AI Containers:

  • Azure provides pre-built containers for vision, speech, language, and text-related tasks.

• Steps for Deployment:

  1. Select the container based on the required AI task (e.g., language detection, sentiment analysis).

  2. Download the container image from Azure Container Registry.

  3. Configure authentication and environment variables.

  4. Deploy and verify functionality locally or in a Kubernetes environment.

5. Advantages of Using Azure AI Services in Containers:

• Offline Capabilities:

  • Run AI models without internet connectivity.

• Hardware Optimization:

  • Optimized for specific hardware configurations for better performance.

• Customizability:

  • Tailor containers for specific tasks (e.g., sentiment analysis, image recognition).

• Scalability and Portability:

  • Scale seamlessly across environments while maintaining portability.

6. Available Azure AI Containers:

• Language Containers:

  • Language detection, named entity recognition, sentiment analysis.

• Speech Containers:

  • Speech-to-text, text-to-speech, and speech translation.

• Vision Containers:

  • Image recognition, object detection, and image analysis.

In this hands-on lecture, we explore deploying Azure AI services as Docker containers to enable scalable and efficient AI solutions. Using a pre-built sentiment analysis container, we demonstrate how to configure, deploy, and test Azure AI services locally or remotely. This session highlights container deployment workflows, configuration of API keys and environment variables, and best practices for managing resources.

Key Points Covered:

1. Introduction to Containerized AI Services:

• Overview of Azure AI services and their pre-built Docker images for tasks like:

  • Sentiment analysis

  • Language detection

  • Computer vision

  • Azure Cognitive Search

• Benefits of containerization:

  • Simplified deployment and scalability

  • Enhanced security with environment isolation

  • Portability across environments

2. Deploying Azure AI Services Containers:

• Container Instance Creation:

  • Navigate to Azure portal and create a new container instance.

  • Select the appropriate Docker image for the desired AI task (e.g., sentiment analysis).

• Configuration:

  • Set resource group, container name, and region.

  • Increase container resources (e.g., 4GB RAM) for optimized performance.

  • Configure networking (e.g., port 5000 for API communication).

  • Add environment variables:

    • API Key: For authentication.

    • Billing Info: For Azure usage tracking.

    • EULA Acceptance: To comply with licensing terms.

3. Testing the Containerized AI Service:

• Deployment Verification:

  • Use Fully Qualified Domain Name (FQDN) of the container instance to access the service.

• Sending API Requests:

  • Utilize curl commands or REST clients to send requests to the container instance.

  • Example input: "The performance was amazing!"

    • Output: Sentiment detected as positive with a confidence score.

• Resource Monitoring:

  • Monitor CPU, memory, and network usage to evaluate container performance.

4. Best Practices for Resource Management:

• Stopping and Deleting Resources:

  • Stop the container instance when not in use to save costs.

  • Delete unnecessary resources, such as alerts and unused containers, to maintain a clean Azure environment.

• Proactive Monitoring:

  • Use Azure Monitor to track container performance metrics for scalability and optimization.

In this lecture, we delve into the concept of content safety and explore how Azure AI services enable you to ensure safe and trustworthy content within your applications. Content safety protects users from harmful material, supports compliance with regulations, and enhances user trust. This session introduces the features, benefits, and use cases of Azure AI services for content safety, preparing you for hands-on implementation in the next lecture.

Key Points Covered:

1. Understanding Content Safety:

• Definition:

  • Ensures only safe, trustworthy content is processed within the Azure Cloud.

  • Protects users from harmful, offensive, or illegal material.

• Purpose:

  • Builds user trust and confidence.

  • Supports ethical AI practices and regulatory compliance.

2. Features of Azure Content Safety:

• Text Moderation:

  • Detects harmful language, offensive speech, and hate content.

• Image Moderation:

  • Identifies explicit imagery or inappropriate visuals.

• Custom Rules:

  • Allows defining organization-specific policies to flag or filter content.

• Automated Filtering:

  • Reduces manual moderation efforts by automatically identifying unsafe content in text, images, or speech.

3. How Content Safety Works:

• Input Data:

  • Content such as text, images, or audio is fed into the system.

• Rule-Based Scanning:

  • Scans for violations based on predefined policies and flags unsafe content.

• Feedback and Improvement:

  • Learns from user feedback to improve moderation accuracy over time.

4. Key Technologies Behind Content Safety:

• Natural Language Processing (NLP):

  • Detects harmful language patterns in text.

• Computer Vision:

  • Analyzes images and videos to identify unsafe content.

• Subdermal Models:

  • Continuously improve accuracy through machine learning.

5. Use Cases for Content Safety:

• Social Media Platforms:

  • Moderates user-generated content and posts.

• Online Forums and Chats:

  • Filters hateful or offensive discussions in real-time.

• E-commerce Sites:

  • Ensures product reviews and comments comply with community guidelines.

• Gaming and Entertainment:

  • Monitors live streams and in-game chats for unsafe behavior.

• Enterprise Use:

  • Enforces company-specific content policies and ensures compliance.

6. Challenges and Solutions:

• Scalability:

  • Handles high data volumes efficiently.

• Accuracy:

  • Combines AI-based detection with human review for improved precision.

• Bias Detection:

  • Requires regular audits to mitigate model bias and ensure fairness.

In this hands-on lecture, we demonstrate how to utilize Azure AI's Content Safety API to moderate and filter harmful, offensive, or inappropriate content. You’ll learn to deploy the Content Safety service, configure access control, and test its functionality using real-world examples. This session highlights text moderation, adjustable severity thresholds, and the flexibility of Azure AI services in managing unsafe content effectively.

Key Points Covered:

1. Setting Up Content Safety Service:

• Deploying the Service:

  • Navigate to the Azure portal and create a new Content Safety API resource.

  • Configure subscription, resource group, and region.

  • Select the appropriate pricing tier (e.g., Standard S0).

• Configuring Access Control:

  • Assign the Cognitive Service User Role to grant appropriate access to users or applications.

  • Ensure roles are propagated to avoid access issues.

2. Exploring Azure AI Studio:

• Using Content Safety APIs:

  • Access Content Safety features via Azure AI Studio.

  • Explore capabilities like text, image, and multimodal moderation.

• Features of Content Safety:

  • Detect harmful, offensive, or inappropriate user-generated and AI-generated content.

  • Leverage APIs for moderation tests and severity adjustments.

3. Moderating Text Content:

• Running Tests:

  • Use sample text inputs to test for offensive, hateful, or violent content.

  • Example inputs include safe text, misspelled violent content, or multiple risk categories.

• Configuring Filters:

  • Adjust severity thresholds for violence, self-harm, sexual content, and hate speech.

  • Define custom rules to block or allow specific content types.

4. Adjusting Severity Thresholds:

• Low, Medium, and High Levels:

  • Define thresholds for different content categories.

  • Example: Block high-severity violent content while allowing low-severity content.

• Testing Threshold Changes:

  • Experiment with threshold adjustments and observe their impact on moderation results.

5. Challenges and Solutions:

• Access Issues:

  • Resolve resource access errors by assigning appropriate roles and permissions.

  • Allow time for access role propagation in Azure.

• Testing Multiple Scenarios:

  • Experiment with various text inputs and severity levels to fine-tune moderation.

In this lecture, we take a deep dive into Azure's Content Safety API, exploring advanced content moderation techniques for text, images, code, and user prompts. Learn how to classify and filter harmful or inappropriate content based on configurable thresholds for various categories, such as violence, self-harm, and sexual content. This session also demonstrates practical use cases and code integration for leveraging the Content Safety API in real-world applications.

Key Points Covered:

1. Advanced Text Moderation:

• Exploring Complex Examples:

  • Analyze and classify text with multiple risk categories (e.g., self-harm and violence).

  • Experiment with real-world scenarios, such as suicide-related content, hateful language, and mixed-language statements.

• Configurable Thresholds:

  • Adjust thresholds for categories like violence, self-harm, and hate speech to block or allow content based on severity.

  • Understand how thresholds impact moderation decisions for high-risk content.

2. Image Moderation:

• Content Classification:

  • Detect and block inappropriate imagery based on self-harm, violence, or sexual content categories.

• Threshold Management:

  • Customize thresholds to allow low-risk content while blocking high-risk material.

• Examples:

  • Test safe images, self-harm scenarios, and AI-generated sexual content to understand how the API applies moderation rules.

3. Moderating User Prompts:

• Prompt Jailbreak Detection:

  • Identify and block user prompts attempting to bypass system rules (e.g., "Do Anything Now" scenarios).

• Ensuring Safety in Interactive Systems:

  • Moderate user-generated inputs in conversational AI or chatbots for compliance and safety.

4. Code Content Moderation:

• Protected Code Detection:

  • Detect code snippets that violate intellectual property rights or licensing terms.

  • Identify matches in public repositories or restricted libraries and flag them for review.

5. Key Features and Functionality:

• Multi-Category Detection:

  • Handle content that spans multiple risk categories simultaneously.

• Detailed Reporting:

  • Gain insights into severity levels, risk classifications, and threshold breaches.

• API Integration:

  • Use prebuilt code templates in Python, Java, or C# to integrate the Content Safety API into applications.

6. Practical Implementation:

• Experimentation and Testing:

  • Utilize sample content to configure and fine-tune thresholds for different content categories.

• Code Integration:

  • Replace placeholders in provided code samples with subscription keys and endpoints to deploy Content Safety APIs in real-world scenarios.

• Resource Management:

  • Safely delete unused resources to optimize Azure costs and maintain a clean environment.

In this lecture, we introduce Azure AI Vision and explore its comprehensive capabilities for image and video analysis. You will learn about tools like Azure Vision Studio and the Computer Vision API, which enable tasks such as image classification, object detection, face analysis, and optical character recognition (OCR). This session provides an overview of the functionalities and sets the foundation for implementing advanced computer vision solutions using Python and Azure AI services.

Key Points Covered:

1. Overview of Azure AI Vision API:

• What is Azure AI Vision?

  • A powerful toolset for analyzing images and videos using Azure AI services.

  • Enables tasks such as tagging, object detection, caption generation, and spatial analysis.

• Features of Azure Vision Studio:

  • A unified interface for exploring computer vision functionalities.

  • Pre-built and customizable tools for image and video analysis.

2. Key Use Cases for Computer Vision Solutions:

• Image Analysis:

  • Generate tags, captions, and detect objects within images.

  • Perform background and foreground removal using the Analyze Image API.

• Image Classification:

  • Leverage pre-built and custom models to classify images.

• Face Detection:

  • Detect and analyze faces within an image, including attributes like age and emotions.

• OCR (Optical Character Recognition):

  • Extract text from printed or handwritten documents using OCR.

• Video Indexing:

  • Analyze video content to extract meaningful insights and metadata.

3. Tools and Resources Used:

• Azure Vision Studio:

  • Explore functionalities like face analysis, spatial analysis, and OCR through an intuitive UI.

• Programming Support:

  • Sample code for Python and C# provided in the designated repository.

  • Hands-on examples for analyzing images and videos programmatically.

4. Key APIs Explored in This Section:

• Analyze Image API:

  • Identify tags, objects, and captions in images.

• Custom Vision API:

  • Create and use custom models for specific classification tasks.

• OCR API:

  • Extract both handwritten and printed text from images.

• Face API:

  • Detect and analyze facial features and attributes.

5. Practical Implementation:

• Using Vision Studio:

  • Navigate and experiment with features like optical character recognition and face analysis.

• Python SDK Integration:

  • Use Python SDK to automate tasks such as image classification and object detection.

• Building Custom Models:

  • Explore the creation of custom vision models for unique use cases.

In this lecture, we explore the diverse functionalities offered by Azure Computer Vision APIs. These powerful tools enable image and video analysis, including object detection, face recognition, optical character recognition (OCR), and more. This session provides a detailed understanding of the capabilities, use cases, and customization options available for implementing computer vision solutions, setting the stage for hands-on demonstrations in the upcoming videos.

Key Points Covered:

1. Image Analysis Using Azure Computer Vision:

• Capabilities:

  • Tagging images, generating captions, and detecting inappropriate content.

  • Extracting metadata and recognizing objects, scenes, and activities.

• Use Cases:

  • Automatic metadata tagging for media libraries.

  • Image search and categorization for organized archives.

2. Custom Vision Models:

• Overview:

  • Create tailored models for specific business needs using your own labeled data.

  • Train, evaluate, and deploy custom models for unique classification tasks.

• Use Cases:

  • E-commerce: Classify product images.

  • Manufacturing: Identify components in industrial settings.

3. Pre-Built Image Classification APIs:

• Capabilities:

  • Automatically categorize images into predefined categories.

  • Leverage ready-to-use pre-built models or train custom models.

• Use Cases:

  • Media archive sorting.

  • Detecting brand logos in marketing campaigns.

4. Object Detection APIs:

• Capabilities:

  • Identify and locate multiple objects within an image using bounding boxes.

• Use Cases:

  • Inventory management: Track specific items.

  • Surveillance: Monitor objects in security footage.

5. Face Detection and Analysis APIs:

• Capabilities:

  • Detect faces and analyze attributes like age, emotion, and gender.

  • Perform face recognition for identity verification.

• Use Cases:

  • Identity verification for secure systems.

  • Emotion detection for customer feedback analysis.

6. Optical Character Recognition (OCR):

• Capabilities:

  • Extract text (printed or handwritten) from images and documents.

  • Support for multiple languages and structured text extraction.

• Use Cases:

  • Digitizing documents and receipts.

  • Automating invoice processing.

7. Video Analysis APIs:

• Capabilities:

  • Process video frames to detect objects, recognize actions, and analyze faces.

  • Monitor movement and recognize activities across frames.

• Use Cases:

  • Video surveillance and event monitoring.

  • Generating highlights for video content.

8. Summary:

• Core Features of Azure AI Vision:

  • Object detection, face recognition, text extraction, and video analysis.

• Applications:

  • Ideal for content moderation, identity verification, document processing, and video analytics.

In this lecture, we dive into the hands-on implementation of Azure AI Vision services to analyze images. You will learn how to use pre-built APIs and Python SDK to extract information such as captions, tags, and objects from images. We will also explore advanced features like background removal, foreground extraction, and detecting people or objects in images. This practical session provides a step-by-step approach to leveraging Azure Vision APIs for real-world applications.

Key Points Covered:

1. Setting Up the Environment:

• Clone the Microsoft-provided Ms. Learn AI Vision GitHub repository.

• Configure the necessary environment variables:

  • AI Service Endpoint

  • Service Key

• Install required Python libraries:

  • matplotlib for visualizing results.

  • Pillow for image manipulation.

  • Azure Vision-specific libraries for API integration.

2. Analyzing Images:

• Image Captioning:

  • Extract captions and dense captions for images with confidence scores.

  • Example:

    • Caption: "Man walking a dog on a leash on a street."

    • Confidence: 82.07%.

• Tagging:

  • Generate descriptive tags for images.

  • Example Tags: "man," "suit," "glasses," "spokesperson."

• Object Detection:

  • Identify objects in images with bounding boxes.

  • Save results as images with highlighted bounding boxes.

3. Advanced Features:

• Background Removal:

  • Remove the background from an image and save the result.

  • Output: Clean, background-free images for further use.

• Foreground Matting:

  • Highlight the primary subject while removing other elements.

4. Hands-On Examples:

• Analyze various images provided in the repository:

  • Street.jpeg - Detect vehicles, pedestrians, and street elements.

  • Person.jpeg - Extract detailed captions and tags related to an individual.

  • Building.jpeg - Identify architectural elements and objects.

• Output:

  • Results Visualization:

    • Captions and tags displayed in the terminal.

    • Enhanced images saved with bounding boxes and processed elements.

5. Practical Applications:

• Generate metadata for media files.

• Detect people and objects in surveillance footage.

• Enhance e-commerce product listings by categorizing and tagging images.

• Remove or isolate background elements for design purposes.

In this lecture, you will learn how to create a custom image classification model using Azure AI Vision services. Unlike pre-built models, this session focuses on building a tailored model for classifying specific categories, such as fruits. You will gain hands-on experience in preparing data, creating a storage account, and setting up the environment to train and deploy the custom model.

This lecture lays the groundwork for training custom models with Azure AI Vision and provides practical insights into real-world applications of image classification.

Key Points Covered:

1. Overview of Custom Image Classification:

• Explanation of the need for custom models to classify specific categories.

• Examples of categories: Fruits (Apple, Banana, Orange).

• Use cases: E-commerce image tagging, inventory management.

2. Setting Up Azure Resources:

• Creating a Computer Vision Resource:

  • Configure resource group, region, and pricing tier.

  • Use the Computer Vision API for building the model.

• Creating a Storage Account:

  • Set up Azure Blob Storage for storing training and test data.

  • Enable anonymous blob access for seamless data retrieval.

3. Preparing Data for Model Training:

• Organizing training and test images into labeled folders.

• Using the provided training_label.json file to define metadata for the images.

• Automating JSON file updates:

  • Editing storage account references in the JSON file using a PowerShell script (replace.ps1).

  • Adjusting PowerShell settings to allow script execution.

4. Uploading Data to Azure Storage:

• Creating a container (e.g., "fruit") in the storage account.

• Configuring appropriate access levels for the container.

• Uploading training images and the JSON metadata file to the container.

5. Preparing for Model Training:

• Connecting the storage account with the Computer Vision resource.

• Verifying uploaded data in the storage container.

• Preparing to train the model using Azure’s Computer Vision APIs.

In this lecture, you will learn how to build a custom image classification model using Azure Vision Studio and Computer Vision resources. Starting with the preparation of training data and its associated metadata in a COCO file, we will proceed to create and train a custom model tailored to classify images into specific categories (e.g., fruits like Apple, Banana, and Orange). This lecture provides a hands-on demonstration of model training, evaluation, and testing, with a focus on real-world use cases.

Key Points Covered:

1. Preparing the Data Set:

• Training Images: Organize images into categories (e.g., Apple, Banana, Orange).

• Metadata File (COCO):

  • Understanding the role of the training_label.json file.

  • Updating storage references in the file using a PowerShell script (replace.ps1).

• Storage Account:

  • Upload training images and COCO file to Azure Blob Storage.

  • Configure the container with appropriate permissions for Azure Vision Studio access.

2. Creating a Custom Model in Azure Vision Studio:

• Accessing Azure Vision Studio:

  • Navigate to the portal and select the configured Computer Vision resource.

• Defining the Dataset:

  • Link the Blob Storage container and COCO file to create a new dataset.

  • Grant Azure Vision Studio the necessary permissions to read from the container.

• Custom Model Setup:

  • Specify model name, type (e.g., Image Classification), and dataset.

  • Allocate training resources with a configurable budget (e.g., 1 hour).

3. Training and Testing the Custom Model:

• Training Process:

  • Initiate model training in Azure Vision Studio.

  • Understand the training status and resource usage.

• Model Testing:

  • Upload test images to evaluate model performance.

  • Receive JSON responses with predictions and confidence scores.

  • Example predictions:

    • Apple: 99.97% confidence.

    • Banana: High confidence.

    • Orange: 99.97% confidence.

4. Post-Training Actions:

• Deleting unused resources:

  • Clean up the Azure storage account and Computer Vision resource to minimize costs.

In this hands-on lab session, we focus on detecting objects in images using Azure Custom Vision, a part of Azure's AI services. Unlike the prebuilt computer vision APIs, we will create and train our own custom object detection model. This lecture guides you step-by-step through the process, from setting up the resources to training and testing your model, and finally cleaning up resources. By the end of this session, you will have a thorough understanding of how to utilize Azure Custom Vision for object detection.

Key Takeaways:

• Introduction to Azure Custom Vision:

  • Overview of the Custom Vision service.

  • Differentiating between classification and object detection models.

• Setting Up Resources:

  • Creating a new Custom Vision project via the Azure portal.

  • Configuring training and prediction resources with the free tier (F0).

• Uploading and Tagging Images:

  • Importing training images with multiple objects in each image.

  • Tagging images manually for object identification (e.g., apple, orange, banana).

• Training the Model:

  • Ensuring minimum image requirements per tag (15 images per object).

  • Running a quick training session and monitoring the iteration process.

• Testing the Model:

  • Using the model for predictions with test images.

  • Understanding confidence levels and threshold adjustments for accurate object detection.

• Performance Metrics:

  • Reviewing precision, recall, and other performance indicators for the trained model.

• Cleaning Up Resources:

  • Deleting Custom Vision projects and associated resources to avoid unnecessary costs.

Lab Workflow:

1. Resource Creation:

   • Access Azure Portal → AI Services → Custom Vision → Create resources for training and prediction.

2. Custom Vision Setup:

   • Log in to the Custom Vision portal.

   • Create a new project for object detection.

3. Image Management:

   • Upload images to the Custom Vision portal.

   • Tag each object in the images accurately.

4. Training the Model:

   • Add sufficient images per tag.

   • Run a training session with a quick training budget.

5. Testing the Model:

   • Upload test images.

   • Verify detection results with confidence scores.

6. Resource Cleanup:

   • Delete Custom Vision projects and Azure resources after completing the lab.

In this lecture, we explore the powerful capabilities of the Azure Computer Vision API and Face API through a hands-on lab. The session is divided into two parts: detecting people in images using the Computer Vision API and analyzing facial features using the Face API. By the end of this lecture, you will understand how to integrate these APIs into your applications for face and people detection, along with extracting key features.

Key Takeaways:

1. Computer Vision API:

   • Detects people in images and draws bounding boxes.

   • Use of Azure.AI.Vision Python library for API integration.

   • Authentication with endpoints and keys stored in environment variables.

   • Analysis of images with specified visual features for detection.

   • Hands-on implementation with Python to detect people and visualize results.

2. Face API:

   • Detects faces and extracts facial features (e.g., age, emotion, gender).

   • Creating and configuring a new Face API resource in Azure.

   • Practical implementation of face detection and feature extraction.

3. Resource Setup and Configuration:

   • Using Azure Portal to create necessary resources (Computer Vision and Face API).

   • Choosing appropriate pricing tiers for development and testing (e.g., F0 free tier).

   • Integration of API keys and endpoints into the project for seamless functionality.

4. Hands-On Implementation:

   • Installing required libraries and modules (Azure.AI.Vision).

   • Writing Python code for people detection using the Computer Vision API.

   • Drawing bounding boxes around detected people in images.

   • Saving and reviewing the output file with bounding boxes.

   • Setting up and coding for face detection and feature analysis using the Face API.

5. Best Practices and Debugging:

   • Ensuring a clean project structure with separate environments (.env files for keys).

   • Activating Python virtual environments for dependency management.

   • Debugging installation issues by specifying library versions when required.

6. Output and Visualization:

   • Generated bounding boxes around detected people in images.

   • Extracted facial features (using the Face API) for further analysis.

Lab Workflow:

1. Resource Creation:

   • Create Computer Vision and Face API resources on Azure Portal.

   • Configure endpoints and keys for API access.

2. Python Setup:

   • Activate a Python virtual environment.

   • Install the required libraries (e.g., Azure.AI.Vision).

   • Set up .env files for storing API keys and endpoints.

3. Implementation:

   • Write and execute code to detect people in images using Computer Vision API.

   • Save the output image with bounding boxes.

   • Extend functionality to detect faces and extract features using Face API.

4. Testing and Output:

   • Test the implemented features with sample images.

   • Visualize bounding boxes for people detection and analyze facial features.

5. Resource Cleanup:

   • Delete created resources in Azure to avoid unnecessary costs.

In this lecture, we dive deeper into the Azure Face API, focusing on detecting faces and extracting detailed facial features from images. The hands-on lab provides practical experience with deploying the Face API resource, integrating it into a Python application, and analyzing various facial attributes such as age, occlusion, emotion, and more. This lecture equips you with the skills to implement face detection and feature extraction in real-world scenarios.

Key Takeaways:

1. Setting Up Face API:

   • Deploying the Face API resource on Azure Portal.

   • Configuring API keys and endpoints in an environment file for secure access.

2. Python Environment Setup:

   • Activating a Python virtual environment.

   • Installing the required library: Azure Cognitive Services Vision Face.

3. Authentication:

   • Creating credentials and a Face Client to interact with the Face API.

4. Face Detection and Feature Extraction:

   • Detecting faces in images with a single command.

   • Retrieving detailed facial features, including:

     • Age: Estimates the age of detected faces.

     • Blur Level: Identifies if the face is blurred.

     • Occlusion: Checks for obstructions like masks or hands.

     • Glasses: Detects the presence of eyeglasses.

     • Emotion: Analyzes emotions such as happiness, sadness, or anger.

5. Visualization with Matplotlib:

   • Drawing bounding boxes around detected faces.

   • Annotating images with extracted attributes.

6. Debugging and Testing:

   • Handling invalid requests by refining feature selection.

   • Experimenting with different images and features for broader understanding.

7. Resource Cleanup:

   • Deleting the Face API resource after completing the lab to optimize costs.

Lab Workflow:

1. Resource Deployment:

   • Navigate to Azure AI Services and create a Face API resource.

   • Save the API key and endpoint in a .env file.

2. Environment Setup:

   • Install the required Python library.

   • Authenticate the Face Client using credentials.

3. Code Implementation:

   • Write Python code for detecting faces and extracting features.

   • Utilize Matplotlib to visualize results.

4. Execution:

   • Run the Python script to detect faces in sample images.

   • Analyze the extracted features and verify results.

5. Debugging:

   • Remove or modify features if invalid requests occur.

   • Retest with default or selective feature sets.

6. Resource Cleanup:

   • Delete the Face API resource from Azure Portal.

In this hands-on lab, we explore Optical Character Recognition (OCR) using the Azure Computer Vision API. This lecture demonstrates how to extract text from images, including printed and handwritten text, using Azure AI services. By the end of this session, you will be equipped to implement OCR solutions that read and process text from images effectively.

Key Takeaways:

1. Introduction to OCR:

   • Understanding the concept of OCR and its practical applications.

   • Using Azure Computer Vision API to extract text from images.

2. Setting Up the Environment:

   • Creating and configuring the Computer Vision resource in Azure Portal.

   • Retrieving API keys and endpoints for integration.

   • Configuring the .env file to securely store credentials.

3. Python Environment Setup:

   • Installing the required library: Azure AI Vision Image Analysis.

   • Activating a Python virtual environment for dependency management.

4. Code Implementation:

   • Writing Python code for:

     • Authenticating the API using keys and endpoints.

     • Analyzing images to extract text using the OCR feature.

   • Processing multiple types of text:

     • Printed text (e.g., Lincoln quote image).

     • Handwritten text (e.g., shopping list image).

5. Drawing Bounding Boxes:

   • Visualizing results with bounding boxes around:

     • Each detected text line.

     • Each detected character or word.

   • Using coordinates to draw precise polygons for better visualization.

6. Debugging and Testing:

   • Handling issues such as indentation errors or invalid results.

   • Iteratively improving the code to handle more specific text detection scenarios.

7. Output Results:

   • Saving extracted text and bounding boxes to image files.

   • Reviewing and validating the extracted text and its visual representation.

Lab Workflow:

1. Resource Setup:

   • Create a Computer Vision resource in Azure Portal.

   • Retrieve and configure the API key and endpoint.

2. Environment Configuration:

   • Install the required Python libraries.

   • Authenticate and initialize the Computer Vision client.

3. Image Analysis:

   • Implement OCR to extract text from sample images (e.g., Lincoln quote, handwritten shopping list).

   • Process results to draw bounding boxes around lines, words, and characters.

4. Visualization:

   • Save processed images with bounding boxes.

   • Display results using Matplotlib or other visualization libraries.

5. Debugging and Refinement:

   • Address errors and refine bounding box visualizations for better clarity.

6. Validation:

   • Test OCR with different images and text formats.

   • Compare extracted text against the original to ensure accuracy.

In this hands-on session, we explore Azure Video Indexer, a powerful tool for extracting insights from videos. This lecture demonstrates how to analyze video content, identify objects, extract captions, detect emotions, and recognize entities. By the end of this session, you will understand how to leverage Azure Video Indexer to gain actionable insights from video data.

Key Takeaways:

1. Introduction to Azure Video Indexer:

   • Overview of Azure Video Indexer and its capabilities.

   • Accessing the Video Indexer platform via videoindexer.ai.

2. Setting Up and Uploading Videos:

   • Logging in with your Azure account.

   • Uploading sample videos for analysis.

   • Configuring upload settings and initiating indexing.

3. Video Analysis:

   • Key features extracted during video analysis:

     • Object Detection: Identifies objects like cell phones, aeroplanes, and books, with timestamps.

     • Topics and Keywords: Categorizes video content into topics like technology and environmentalism, associating relevant keywords.

     • Named Entities: Recognizes entities such as individuals, companies, or brands (e.g., "Microsoft" and "Eric").

     • Emotions: Detects emotions like joy and assigns confidence levels.

     • Labels: Tags scenes with descriptive labels.

     • Scenes and Shots: Identifies scenes, shots, and keyframes for detailed segmentation.

4. Visualizing Results:

   • Timeline-based insights:

     • View object appearances with precise timestamps.

     • Analyze topics and keyword coverage over specific intervals.

   • Caption extraction for accessibility and context understanding.

5. Hands-On Demonstration:

   • Uploading and indexing a sample video.

   • Navigating through analysis results (e.g., objects, entities, emotions).

   • Reviewing captions and timeline-based content.

6. Advanced Features:

   • Editing indexed content with the built-in editor.

   • Re-indexing videos with updated settings for improved insights.

7. Resource Cleanup:

   • Deleting analyzed videos to manage storage and costs.

Lab Workflow:

1. Access Video Indexer:

   • Log in to videoindexer.ai using your Azure account.

2. Upload Video:

   • Choose a sample video file (e.g., "ResponsibleAI.mp4").

   • Configure settings and initiate the upload and indexing process.

3. Analyze Video:

   • Explore the insights dashboard.

   • Review objects, entities, and emotions detected in the video.

   • Inspect topics and keywords across the timeline.

4. Extract Insights:

   • Examine captions and timelines.

   • Navigate through scenes, shots, and keyframes.

5. Cleanup:

   • Delete indexed videos to free up resources.

In this hands-on lecture, we explore how to create and train a Custom Image Classification Model using Azure Custom Vision. This session demonstrates the process of preparing a dataset, configuring Azure services, training a model, and testing the results. By the end of this lecture, you will have a comprehensive understanding of building and deploying a custom classification model tailored to your specific data.

Key Takeaways:

1. Introduction to Custom Image Classification:

   • Overview of custom image classification and its applications.

   • Leveraging Azure Custom Vision for training and prediction.

2. Dataset Preparation:

   • Using a dataset with two classes: dogs and cats.

   • Uploading images with proper tagging for each class.

3. Azure Custom Vision Setup:

   • Creating a Custom Vision resource in Azure for training and prediction.

   • Choosing the F0 free tier for cost-effective development.

   • Configuring the resource group and region for deployment.

4. Project Configuration:

   • Setting up a new project in the Custom Vision portal.

   • Defining the classification task as multi-class (single tag per image).

   • Uploading and tagging images for training.

5. Model Training:

   • Initiating quick training with a minimum budget.

   • Adjusting probability thresholds for classification accuracy.

   • Publishing the trained model for prediction via an API.

6. Testing the Model:

   • Using the Quick Test feature to evaluate the model.

   • Testing with unseen images for both dog and cat classes.

   • Reviewing prediction probabilities and fine-tuning if needed.

7. Resource Cleanup:

   • Unpublishing iterations and deleting the project.

   • Ensuring all resources (training and prediction) are deleted to manage costs.

Lab Workflow:

1. Azure Resource Creation:

   • Navigate to Azure Portal → AI Services → Custom Vision.

   • Create resources for training and prediction.

2. Project Setup:

   • Log in to the Custom Vision portal.

   • Create a project for classification and upload datasets.

   • Tag images accurately for training purposes.

3. Training the Model:

   • Initiate quick training with a minimum threshold.

   • Monitor the training process and adjust parameters as needed.

4. Testing the Model:

   • Use the Quick Test option to evaluate performance.

   • Test with different images and review prediction results.

5. Cleanup:

   • Unpublish iterations and delete the Custom Vision project.

   • Remove all associated Azure resources.

In this lecture, we explore the pricing structure of Azure Computer Vision API and related AI services. Using the Azure Pricing Calculator, we estimate costs for various operations such as image analysis, face detection, and video indexing. By the end of this lecture, you will have a solid understanding of how to estimate Azure AI service costs based on usage, helping you optimize budgets for real-world projects.

Key Takeaways:

1. Introduction to Azure Computer Vision Pricing:

   • Overview of pricing models for Azure Computer Vision API.

   • Understanding free tier (F0) and standard pricing (S1).

2. Using the Azure Pricing Calculator:

   • Accessing the Azure Pricing Calculator via the browser.

   • Estimating costs based on transaction volume and resource usage.

3. Detailed Cost Analysis:

   • Group 1 Operations:

     • Examples: Tag extraction, image analysis.

     • Cost: $1 per 1,000 transactions for the first 1 million transactions.

   • Group 2 Operations:

     • Examples: Dense captions, object detection.

     • Cost: $1.50 per 1,000 transactions.

4. Cost Reduction with High Usage:

   • Discounts for higher transaction volumes:

     • 10M–100M transactions: $0.60 per 1,000 transactions.

     • Beyond 100M: Further reduced rates.

5. Specific Pricing Examples:

   • Face API:

     • $1 per 1,000 transactions for the first 1M transactions.

     • Discounts applied for higher usage.

   • Custom Vision:

     • Free tier: 2 projects, 1 hour of training/month, 5,000 images free.

     • Standard pricing: $10/hour for training, $2 per 1,000 predictions.

   • Video Indexer:

     • Basic insights: $0.06 per minute.

     • Standard insights: $0.11 per minute.

     • Advanced insights: $0.19 per minute.

6. Region-Specific Pricing:

   • Pricing variations across regions (e.g., East US vs. Australia).

   • Importance of comparing costs for optimal region selection.

7. Practical Insights for Architects:

   • Understanding how transaction volumes and feature requirements impact costs.

   • Using the calculator for detailed budget planning.

Lab Workflow:

1. Accessing the Azure Pricing Calculator:

   • Navigate to the calculator online.

   • Select Azure AI services like Computer Vision, Custom Vision, and Video Indexer.

2. Configuring Usage Scenarios:

   • Input expected usage (e.g., number of transactions, training hours).

   • Review cost estimates for free and standard tiers.

3. Analyzing Pricing:

   • Compare group-specific costs for Computer Vision operations.

   • Evaluate total costs for large-scale usage with tiered discounts.

4. Adjusting Parameters:

   • Modify usage inputs (e.g., region, transaction count).

   • Reassess cost implications based on updated inputs.

5. Interpreting Results:

   • Break down costs for each operation type.

   • Generate reports for organizational budget discussions.

In this lecture, we introduce Azure AI Natural Language Processing (NLP) APIs and explore their capabilities for analyzing text, creating custom models, and performing advanced language tasks. This session provides an overview of Azure Language Studio and its tools, empowering learners to extract insights, build custom language models, and perform text and speech translation. By the end of this session, you'll have a clear understanding of Azure's NLP capabilities and how to use them in real-world applications.

Key Takeaways:

1. Overview of Azure AI NLP APIs:

   • Language Services: Analyze and process text for insights.

   • Speech Services: Enable speech-to-text, text-to-speech, and speech translation.

2. Text Analysis Features:

   • Detect language and analyze sentiment.

   • Extract named entities (e.g., people, places, organizations).

   • Identify key phrases for summarization and insights.

   • Use prebuilt tasks for specific needs (e.g., post-call transcription).

3. Custom NLP Models:

   • Build custom text classification and entity extraction models.

   • Leverage Azure Language Studio for creating and managing custom models.

   • Use Microsoft-provided sample data for training models via GitHub.

4. Translation Services:

   • Perform text-to-text translation (e.g., English to French).

   • Translate speech-to-speech while preserving linguistic nuances:

     • Speech transcribed into text.

     • Text translated into the target language.

     • Translated text synthesized back into speech.

5. Prebuilt vs. Custom Models:

   • Prebuilt models:

     • Ready-to-use for common tasks like named entity recognition and key phrase extraction.

   • Custom models:

     • Tailored to specific business needs using labeled training data.

6. Azure Language Studio:

   • Centralized tool for managing NLP tasks.

   • Supports both prebuilt tasks and custom model development.

   • Provides APIs for seamless integration into applications.

Lab Workflow:

1. Setting Up Language Studio:

   • Access Azure Language Studio via the Azure Portal.

   • Explore prebuilt tasks like sentiment analysis and named entity recognition.

2. Text Analysis:

   • Use the Analyze Text API to extract language, sentiment, and entities.

   • Identify key phrases from input text for summarization.

3. Custom Model Development:

   • Create a custom text classification model using labeled data.

   • Train a custom named entity recognition (NER) model.

4. Translation and Speech Synthesis:

   • Translate text between languages using Azure Translator.

   • Perform speech-to-speech translation with language and speech services.

5. Testing and Validation:

   • Test prebuilt and custom models with sample data.

   • Validate results and refine models as needed.

This lecture provides an in-depth overview of Natural Language Processing (NLP) solutions offered by Azure AI Services, covering a variety of APIs for text analysis, language understanding, translation, and speech synthesis. The session highlights use cases, features, and practical applications of these APIs, enabling learners to understand how Azure AI empowers NLP tasks. By the end of this lecture, learners will have a clear understanding of Azure NLP capabilities and their integration into real-world projects.

Key Takeaways:

1. Text Analysis with Azure AI Language API:

   • Extract meaningful insights from text using Text Analytics API:

     • Sentiment Analysis: Assess the emotional tone of text.

     • Language Detection: Identify the language of the input text.

     • Key Phrase Extraction: Summarize text content.

     • Named Entity Recognition (NER): Recognize entities like names, dates, and locations.

   • Use Cases:

     • Social media monitoring.

     • Customer feedback analysis.

2. Question Answering with Azure Language:

   • Build intelligent Q&A systems using prebuilt models.

   • Create knowledge bases with Azure Cognitive Search.

   • Use Cases:

     • Customer support bots.

     • FAQ systems.

3. Conversation Language Understanding:

   • Develop models that understand and respond to natural language using Language Understanding API (LUIS):

     • Define intents, entities, and actions for user commands.

     • Train and deploy models for real-time use.

   • Use Cases:

     • Interactive voice response (IVR) systems.

     • Virtual assistants.

4. Custom Text Classification:

   • Classify text into predefined categories based on content.

   • Train custom models using labeled datasets (e.g., product reviews or news articles).

   • Use Cases:

     • Email classification.

     • Sentiment analysis.

5. Custom Named Entity Recognition (NER):

   • Create tailored NER models for specific entities (e.g., organizations, locations, or people).

   • Use prebuilt or custom NER models for extracting valuable insights.

   • Use Cases:

     • Legal document analysis.

     • News categorization.

6. Azure Translator Services:

   • Translate text across 70+ languages in real time.

   • Integrate translation APIs for:

     • Website localization: Adapt content for different languages.

     • Cross-border communication: Facilitate multilingual collaboration.

   • Use Cases:

     • Website translation.

     • Multilingual chat systems.

7. Speech-Enabled Applications:

   • Integrate Speech-to-Text and Text-to-Speech into applications.

   • Use custom voice models for domain-specific needs.

   • Combine translation services with speech APIs for:

     • Real-time conference translation.

     • Accessibility features in apps.

Lab Workflow:

1. Text Analysis:

   • Use Text Analytics API to extract language, sentiment, and key phrases.

   • Apply Named Entity Recognition to identify entities in text.

2. Custom Model Development:

   • Train and deploy models for text classification and custom NER.

   • Use Azure Language Studio for managing models.

3. Translation Services:

   • Translate text between languages using Azure Translator.

   • Implement website localization with APIs.

4. Speech Services:

   • Convert speech to text and back using Azure Speech Services.

   • Combine speech and translation APIs for real-time multilingual conversations.

In this lecture, we explore Azure Language Services and their capabilities for analyzing text. This hands-on session demonstrates how to detect language, analyze sentiment, extract key phrases, recognize entities, and retrieve linked entities using Azure AI. By the end of this lecture, learners will understand how to leverage Azure's Text Analytics API to extract meaningful insights from text data.

Key Takeaways:

1. Introduction to Azure Language Services:

   • Overview of features provided by Azure's Text Analytics API.

   • Use cases include sentiment analysis, language detection, and entity recognition.

2. Text Analysis Features:

   • Language Detection:

     • Identify the language of text (e.g., English, French).

   • Sentiment Analysis:

     • Determine the emotional tone of text (e.g., positive, negative, mixed).

   • Key Phrase Extraction:

     • Summarize text by identifying key phrases.

   • Entity Recognition:

     • Extract entities such as names, locations, dates, and organizations.

   • Linked Entity Recognition:

     • Identify entities and link them to relevant internet resources (e.g., Wikipedia).

3. Azure Language Resource Setup:

   • Navigate to the Azure Portal and create a Language Services resource.

   • Configure API keys and endpoints for integration.

4. Practical Implementation:

   • Clone a prebuilt GitHub repository with Python scripts for text analysis.

   • Configure .env files with API keys and endpoints.

   • Execute Python scripts to perform text analysis tasks.

5. Hands-On Demonstration:

   • Analyze sample text reviews using Python and Azure Text Analytics API:

     • Detect language for text reviews.

     • Perform sentiment analysis for positive, negative, and mixed reviews.

     • Extract key phrases for summarization.

     • Recognize entities like dates, locations, and organizations.

     • Retrieve linked entities with external references.

6. Output Validation:

   • Review and interpret analysis results:

     • Sentiment scores and associated language.

     • Extracted phrases and entities.

     • Backlinks for additional information about entities.

Lab Workflow:

1. Azure Resource Creation:

   • Create a Language Services resource in Azure.

   • Configure API keys and endpoint.

2. Environment Setup:

   • Clone the GitHub repository for text analysis.

   • Install required libraries (e.g., Azure AI Text Analytics).

3. Text Analysis Execution:

   • Run Python scripts to analyze sample text files.

   • Implement:

     • Language detection.

     • Sentiment analysis.

     • Key phrase extraction.

     • Entity recognition.

     • Linked entity recognition.

4. Result Analysis:

   • Validate outputs for each text analysis feature.

   • Interpret results for practical use cases.

In this lecture, we demonstrate how to create a Language Understanding (LUIS) model using Azure Language Studio. The session focuses on building a conversational banking chatbot by defining intents and entities and mapping user queries to specific actions. By the end of this lecture, learners will understand how to develop, train, and deploy a language understanding model for conversational AI applications.

Key Takeaways:

1. Introduction to Language Understanding Models:

   • Language Understanding enables mapping user input in natural language to predefined intents and entities.

   • Use case: Building intelligent chatbots and virtual assistants.

2. Key Components of a Language Understanding Model:

   • Intents: Represent the user's goal or purpose.

     • Example: "Check account balance" or "Transfer money."

   • Entities: Extract specific details from user input.

     • Example: Amount, date, or account type.

3. Hands-On Demo: Creating a Banking Chatbot:

   • Scenario: A chatbot that assists with banking services like balance inquiry, money transfer, and recent transactions.

   • Example Intents:

     • "Check account balance."

     • "Transfer money."

     • "View recent transactions."

   • Example Entities:

     • Amount: e.g., $100.

     • Account type: e.g., "savings account."

     • Date: e.g., "yesterday."

4. Azure Language Studio Setup:

   • Navigate to Azure Language Studio and create a new project.

   • Configure project settings:

     • Name: "Banking Bot."

     • Language: English.

     • Resource: Select the Azure Language Service (free tier).

5. Data Preparation and Model Configuration:

   • Define Intents:

     • Add intents such as "Check account balance" and "Transfer money."

   • Add Entities:

     • Configure entities for extracting details like amount, date, and account type.

   • Sample Utterances:

     • Provide sample user queries for each intent.

     • Annotate utterances to associate entities with specific parts of the text.

6. Training the Model:

   • Input sufficient sample data to ensure robust model performance.

   • Train the model to recognize user intents and extract entities accurately.

7. Next Steps:

   • Deploy the trained model for testing and real-time use.

   • Test the chatbot with various queries to validate accuracy.

Lab Workflow:

1. Setup:

   • Access Azure Language Studio and sign in with your Azure account.

   • Create a new project for language understanding.

2. Define Model Components:

   • Add intents for user actions (e.g., "Check balance").

   • Define entities for specific details (e.g., amount, date).

3. Input Sample Data:

   • Provide at least five sample utterances per intent.

   • Annotate sample queries to tag entities.

4. Train the Model:

   • Run a training job to improve model performance.

   • Review and refine the model based on feedback.

5. Deploy and Test:

   • Deploy the model to make it accessible via APIs.

   • Test the model using sample queries and validate results.

In this lecture, we take a step-by-step approach to creating, training, and testing a chatbot using Azure Language Studio. By the end of the session, you'll have a clear understanding of how to manage intents, define entities, and evaluate a chatbot's performance for real-world scenarios.

Key Topics Covered:

1. Defining Intents and Entities

   • Adding multiple entries (e.g., "Check Account Balance," "Transfer Money").

   • Understanding how entities (e.g., date, amount) enhance the chatbot's ability to extract actionable data.

2. Saving and Organizing Data

   • Common pitfalls such as forgetting to save changes.

   • Filtering data based on specific intents (e.g., only showing queries related to account balance).

3. Tagging Data for Improved Understanding

   • Tagging user inputs as entities (e.g., "checking account" as an account type, "$150" as an amount).

   • Using regular expressions to identify patterns like dates in user queries.

4. Training and Evaluation

   • Training the chatbot with Azure Language Studio:

     • Using 80% of data for training and 20% for testing.

     • Insights into training completion time and progress.

   • Evaluating model performance:

     • Metrics like F1 Score, Precision, and Recall.

     • Addressing feedback on low data intents for better performance.

5. Deploying the Chatbot

   • Deploying the trained model in a production environment.

   • Testing the chatbot with real queries:

     • Examples include “Can I check how much money I have?” and “Show me the last transaction.”

6. Debugging and Optimization

   • Reviewing JSON responses for detailed insights.

   • Refining intents and entities based on test outcomes for enhanced accuracy.

7. Practical Applications

   • Integrating the chatbot with real-world tasks like:

     • Transferring money.

     • Fetching user account details.

In this lecture, you will learn how to perform Custom Text Classification using Azure Language Studio. This hands-on lab demonstrates how to categorize text into multiple labels, create a classification project, and manage the associated Azure resources. By the end of this session, you’ll be equipped to classify textual data effectively using Azure AI tools.

Key Topics Covered:

1. Introduction to Custom Text Classification

   • Overview of text classification and its practical applications.

   • Setting up a classification model in Azure Language Studio.

2. Setting Up Azure Resources

   • Creating and configuring a Storage Account for storing datasets.

   • Enabling Blob Anonymous Access and assigning roles for seamless integration.

3. Preparing the Dataset

   • Downloading and organizing articles into predefined categories (e.g., sports, news).

   • Uploading the dataset to Azure Storage for use in the classification project.

4. Creating a Custom Text Classification Project

   • Step-by-step guide to initiating a Single-Label Classification Project:

     • Naming the project (e.g., “Classify Articles”).

     • Configuring project language and storage settings.

   • Common errors during setup and troubleshooting tips.

5. Labeling the Data

   • Introduction to data labeling for better model accuracy.

   • Challenges in project creation and API limitations.

6. Next Steps and Continuation

   • Overview of the next phase: Labeling data and training the classification model.

   • Addressing temporary API issues to ensure project continuity.

In this lecture, we dive into the process of tagging, training, and deploying a custom text classification model using Azure Language Studio. You will learn how to categorize text into predefined classes, manually label data, train the model, and evaluate its performance. By the end of the session, you’ll have a comprehensive understanding of how to create, deploy, and test a text classification model.

Key Topics Covered:

1. Setting Up Classification Categories

   • Defining classes such as Sports, News, Entertainment, and Classified.

   • Assigning articles to respective categories manually.

2. Data Labeling

   • Manual data labeling for enhanced model accuracy.

   • Overview of auto-labeling functionality (though not used in this example).

3. Training the Model

   • Initiating the training process and monitoring job progress.

   • Insights into training metrics:

     • F1 Score, Precision, and Recall (achieving 100% in this example).

   • Explanation of evaluation steps and their importance.

4. Deploying the Trained Model

   • Deploying the classification model in a production environment.

   • Selecting deployment resources (e.g., region and resource group).

5. Testing and Validation

   • Testing the deployed model with various text samples.

   • Analyzing model predictions, confidence scores, and potential improvements.

6. Python Integration for Classification

   • Configuring Python scripts for text classification:

     • Setting up .env files with service endpoint and key.

     • Using Microsoft’s documentation to integrate predictions into your applications.

7. Cleanup Process

   • Deleting unused projects and Azure resources (e.g., storage accounts) to optimize costs.

In this lecture, you will learn how to deploy a custom Named Entity Recognition (NER) model in Azure Language Studio. The session covers setting up resources, configuring storage, creating a custom NER project, and handling common issues during deployment. By the end of the lecture, you’ll be equipped with the knowledge to build and deploy a custom NER model to extract user-defined entities such as dates, numbers, places, and people from text.

Key Topics Covered:

1. Introduction to Custom Named Entity Recognition (NER)

   • Overview of NER and its applications.

   • Understanding how to define and extract custom entities (e.g., dates, prices, locations).

2. Setting Up Azure Resources

   • Creating a storage account and enabling blob anonymous access for external accessibility.

   • Assigning the Storage Blob Data Contributor Role for permissions.

   • Configuring Azure Language Studio to connect with the correct storage account.

3. Data Preparation

   • Preparing the dataset (e.g., classified ads with entities such as prices, items for sale, and locations).

   • Uploading data to a storage container for use in the NER project.

4. Creating and Configuring a Custom NER Project

   • Setting up a project in Azure Language Studio:

     • Selecting Custom Named Entity Recognition.

     • Defining primary language settings (e.g., English).

   • Associating the project with the correct Azure resources and storage account.

5. Troubleshooting Resource and Configuration Issues

   • Steps to resolve common challenges, such as:

     • Changing storage accounts when the wrong one is pre-selected.

     • Deploying new Azure Language resources to link with updated storage accounts.

6. Next Steps for Custom NER Model

   • Highlighting the process of creating the NER model:

     • Defining entities and labeling data.

     • Training and deploying the model.

   • Overview of issues like resource deployment delays and their impact on project progress.

In this lecture, you will explore the complete process of creating and deploying a Custom Named Entity Recognition (NER) model in Azure Language Studio. This session covers data labeling, model training, deployment, and testing to extract user-defined entities like prices, items for sale, and locations from text. By the end, you’ll have a solid understanding of how to build, train, and deploy custom NER models effectively.

Key Topics Covered:

1. Setting Up the NER Project

   • Creating a Custom Named Entity Recognition (NER) project in Azure Language Studio.

   • Defining custom entities such as:

     • Item for Sale

     • Price

     • Location

2. Data Labeling for Custom Entities

   • Manually tagging entities in documents (e.g., $90 as price, Denver as location).

   • Importance of saving changes after each document to prevent data loss.

   • Optional use of auto-labeling features to simplify the process.

3. Training the Custom NER Model

   • Splitting the dataset (80% for training, 20% for testing).

   • Monitoring training progress and understanding model performance metrics:

     • F1 Score

     • Precision

     • Recall

4. Deploying the NER Model

   • Deploying the trained model to production resources.

   • Associating the deployment with the correct Azure resources.

5. Testing the Deployed Model

   • Testing the model with example text to validate entity extraction:

     • Example: "Bluetooth is for sale at $100 in Denver."

       • Extracted entities:

         • Item for Sale: Bluetooth

         • Price: $100

         • Location: Denver

6. Cleanup and Resource Management

   • Deleting unused projects and resources, including storage accounts and deployments, to manage costs efficiently.

In this lecture, we explore how to set up and implement Azure Language Translation Services using both the Azure Portal and Python code. This session demonstrates the creation of a translator resource, configuration of API keys, and practical testing of translation services across multiple languages through both the browser interface and sample Python code.

Key Topics Covered:

1. Creating an Azure Translator Service

   • Navigating to Azure AI services and creating a Translator service.

   • Setting up the resource with appropriate region and pricing tier.

2. Testing Translation via the Azure Portal

   • Using the web interface to test text translations.

   • Examples of translations from English to Spanish, Gujarati, Hebrew, and Hindi.

   • Auto-detection of source language and the ability to manually change target languages.

3. Accessing Keys and Endpoints

   • Retrieving the API key and endpoint URL for integration.

   • Explanation of how REST APIs enable translation functionality.

4. Implementing Translation via Python Code

   • Using the provided Python sample code to implement language translation.

   • Modifying the code:

     • Adding the correct API key and location (e.g., East US).

     • Fixing common syntax issues in sample code.

   • Running the Python script to translate sample text into French and Zulu.

5. Support for Other Languages and Programming Platforms

   • Availability of C#, Node.js, Java, and Go sample codes for broader implementation.

   • Explanation of integrated multi-language support via the Azure Portal interface.

In this hands-on lecture, we explore how to use Azure Speech Services to implement speech recognition and speech synthesis. You will learn to process audio inputs, transcribe speech into text, and synthesize speech output in a specific voice or language. This session demonstrates the setup, coding, and testing of speech services through Azure and Python.

Key Topics Covered:

1. Introduction to Azure Speech Services

   • Overview of speech recognition and synthesis capabilities.

   • Use cases for converting audio inputs to text and generating speech from text.

2. Setting Up Speech Services in Azure

   • Creating a Speech Services resource in the Azure Portal.

   • Configuring the resource with the correct region and accessing keys and endpoints for integration.

3. Configuring the Development Environment

   • Setting up a Python virtual environment.

   • Installing required modules:

     • azure-cognitiveservices-speech

     • playsound

   • Configuring environment variables with API keys and region.

4. Implementing Speech Recognition

   • Writing Python code to process .wav audio files.

   • Using the Speech SDK to transcribe speech input into text.

   • Handling audio playback with playsound.

5. Implementing Speech Synthesis

   • Generating spoken output from recognized text using Speech Synthesis API.

   • Testing the output with examples like reading the current time in text and voice.

6. Testing the Full Flow

   • Running the application to:

     • Recognize and transcribe audio from a file.

     • Synthesize and playback speech output based on transcribed text.

   • Exploring use cases for real-time input via microphones.

7. Debugging and Error Handling

   • Resolving common issues such as:

     • Missing module installations.

     • Syntax errors in sample code.

   • Ignoring minor warnings that do not impact functionality.

In this lecture, we enhance our speech recognition and synthesis implementation by taking real-time audio input from a microphone. Building upon the previous session, we demonstrate how to configure Azure Speech Services to recognize speech directly from a microphone, process it, and synthesize responses. This hands-on approach illustrates the flexibility of Azure's Speech SDK for dynamic and interactive use cases.

Key Topics Covered:

1. Transition from File-Based to Microphone Input

   • Updating the code to replace static .wav file input with live microphone input.

   • Maintaining existing speech processing logic for real-time interaction.

2. Configuring Microphone Input

   • Using Speech SDK to enable microphone access.

   • Handling live input and passing it to the speech recognition process.

3. Real-Time Speech Recognition

   • Capturing spoken input through the microphone.

   • Recognizing phrases like "What time is it?" or "Hello" and processing them dynamically.

4. Speech Synthesis in Real-Time

   • Generating synthesized speech responses based on recognized input.

   • Examples include responding with the current time or other predefined outputs.

5. Debugging Real-Time Issues

   • Handling issues where recognized input does not trigger the appropriate response.

   • Ensuring logic paths (e.g., tell time) are correctly executed.

6. Testing the Full Interactive Flow

   • Running the program to validate the seamless integration of speech recognition and synthesis.

   • Examples of dynamic interactions, such as asking "What time is it?" or simply saying "Hello."

In this hands-on lecture, we explore speech translation and synthesis using Azure Speech Services. Unlike simple text translation, this session demonstrates how to process speech signals from an audio file or microphone, transcribe them into text, translate the text into multiple languages, and synthesize the translated text back into speech in the target language. By the end of the lecture, you'll have a clear understanding of how to implement multilingual speech translation.

Key Topics Covered:

1. Introduction to Speech Translation

   • Overview of converting speech in one language to text in another.

   • Synthesizing translated text into audio in the target language.

2. Setting Up Azure Speech Services

   • Using previously created Speech Services resources.

   • Configuring API keys and region settings for integration with the service.

3. Data Preparation

   • Using an audio file (station.wav) containing speech input ("Where is the station?").

   • Setting up the Python environment and dependencies (e.g., pygame for audio playback).

4. Implementing Speech Translation and Synthesis

   • Translating speech input from a file into multiple languages (e.g., French, Spanish, Hindi).

   • Synthesizing translated text into speech in the target language.

5. Real-Time Speech Translation via Microphone

   • Modifying the code to accept live audio input from the microphone.

   • Translating phrases spoken into the microphone into the desired target language.

   • Synthesizing and playing back the translated audio.

6. Testing the Complete Workflow

   • Example inputs and outputs:

     • Translating "Where is the station?" into French, Spanish, and Hindi.

     • Translating and synthesizing "I love you" into Hindi using live microphone input.

7. Resource Cleanup

   • Importance of deleting Azure resources after testing to manage costs efficiently.

In this lecture, we explore the pricing structure of Azure Natural Language Processing (NLP) APIs. This session covers the pricing models, free-tier options, and cost estimations for various NLP-related tasks, enabling you to make informed decisions about resource allocation and budgeting for your Azure projects.

Key Topics Covered:

1. Overview of NLP Pricing Models

   • Freemium Model: Free-tier offerings such as 5,000 text records for NLP tasks.

   • Pay-As-You-Go Model: Costs associated with additional usage beyond the free tier.

2. Exploring the Azure Pricing Calculator

   • Navigating the Azure Pricing Calculator to estimate costs.

   • Selecting Azure AI Services to calculate pricing for specific NLP tasks.

3. Task-Specific Pricing Examples

   • Standard NLP Tasks:

     • Sentiment Analysis, Key Phrase Extraction, Named Entity Recognition, and PII Detection:

       • $1 per 1,000 text records per month.

     • Summarization:

       • $2 per 1,000 text records.

   • Custom NLP Models:

     • Custom Named Entity Recognition:

       • Training: $3 per hour.

       • Inference: $0.50 per model per hour.

4. Speech and Translation Services

   • Text Translation:

     • Free-tier: 2 million characters per month.

     • Standard (S1 Model): $10 per 1 million characters.

   • Document Translation:

     • $15 per 1 million characters.

   • Speech Services:

     • Free-tier: 5 hours of audio per month.

5. Key Considerations

   • Costs are subject to change based on Microsoft policies.

   • Using the Azure Pricing Calculator for up-to-date cost estimation and planning.