AI-103: Azure AI Apps & Agents Developer Ultimate Course

Explore the three core abilities of artificial intelligence—learning, reasoning, and self-correction—and how generative AI, NLP, computer vision, computer speech, and information extraction enable intelligent machines.

Microsoft Azure AI Foundry serves as a unified platform to build and manage end-to-end AI applications with models, agents, tools, and knowledge, aligning with the AI 103 certification.

Explore foundry tools, offering pre-built AI capabilities accessed via API or SDK for text analysis, speech, translation, document intelligence, and multi-modal content understanding.

Discover developer tools and SDKs for Azure AI Foundry, including Visual Studio Code, the Foundry extension, and GitHub Copilot, to write, test, and deploy models, agents, and knowledge bases.

Develop responsible ai practices by applying fairness, reliability, transparency, privacy, security, and accountability to build safe, trustworthy systems, recognizing bias in hiring and autonomous vehicle decisions.

Explore the model catalog in Azure AI Foundry to discover, compare, and select Azure and partner models, including large-language and small-language options, using filters and model cards.

Use objective benchmarks in the Azure AI Foundry portal, including quality, safety, cost, and performance, to drive data-driven model selection and apply scenario-based recommendations.

Deploy models to endpoints using Azure AI Foundry, test via playground, and manage deployments with default or custom settings, token per minute rate limits, and deployment names.

Evaluate model performance using manual and automated methods in Azure AI Foundry, including interactive testing and structured reviews. Drive quality assurance, user satisfaction, continuous improvement, and safety.

Explore the model playground to test prompts and models in a sandbox, generate sample code, validate ideas quickly in the Azure AI Foundry portal, and then integrate into your app.

Choose the right endpoint and SDK for Azure AI apps, between Foundry project endpoints or Azure OpenAI endpoints, and select authentication method and response API or chat completion.

Explore how the Foundry response API unifies chat and assistant capabilities for stateful, multi-turn conversations, with easy integration via Foundry sdk or OpenAI sdk.

Explore generating responses with the chat completions API, using system and user prompts in a conversational loop to maintain context for legacy applications or cross-platform compatibility.

Explore how tools extend ai models by fetching real-time data, performing actions, and connecting with external systems through code interpreter, web search, file search, and function calling.

Explore how the code interpreter provides a Python runtime to write and run code, enabling dynamic execution, file handling (csv, json, images), data analysis, math calculations, and real-time feedback.

Learn how the web search tool enables the AI model to access live internet information in real time, improving accuracy and reducing hallucinations. See the workflow from query to result.

Explore the file search tool to retrieve information from uploaded documents using semantic retrieval and vector stores, producing document-grounded answers with citations for enterprise knowledge.

Enable the AI model to call structured functions defined in your app, letting the app execute logic securely, with validation, and return results for live-grounded interactions.

Learn how prompt engineering designs better prompts to elicit accurate and relevant AI responses, by using system, user, and assistant messages and practical examples.

Explore prompt design techniques—persona, format templates, chain-of-thought, and few-shot learning—and learn how temperature and top-p settings shape response quality for Azure AI apps.

Explore retrieval augmented generation and grounding to improve accuracy with external data; learn data ingestion, including OCR and PDF extraction, embeddings, vector databases, semantic search, and context-driven LLM responses.

Fine-tune pre-trained language models on domain-specific data to improve task performance, using supervised, reinforcement, or direct preference optimization techniques.

Demonstrates fine-tuning a GPT-4.1 model in the Microsoft Foundry portal using supervised training, uploading training data, deploying the fine-tuned model, and testing it in the playground to train domain-specific knowledge.

Compare prompt engineering, rad, and fine-tuning to weigh time, cost, and complexity, highlighting best uses for tone, formatting, factual accuracy with azure ai search and azure storage, and behavioral consistency.

Identify potential harms in responsible AI solutions, prioritize risks by impact and likelihood, test with Red Team simulations, and document findings for stakeholder risk mitigation.

Identify potential risks in AI systems and measure them by creating test prompts that trigger harmful responses. Review outputs and evaluate harm to establish a baseline and assess safety improvements.

Mitigate AI risks with a layered protection approach across model, safety system, grounding, and user experience layers, combining model tuning, guardrails, system prompts, and RAG-based grounding.

Learn to implement and customize guardrails in Microsoft Foundry, configuring built-in and custom controls to block harmful prompts and apply risk-based blocking to GPT-4.1.

Explore what ai agents are and how they automate tasks with Microsoft Foundry Agent Service, understanding context, taking actions, and integrating tools for smarter, scalable business workflows.

Discover Microsoft Foundry Agent Service, a platform that simplifies building, deploying, and scaling AI agents with prompts and tools, offering declarative and hosted agents with secure data handling and observability.

Compare the two common agent development approaches in Microsoft Foundry: the portal’s rapid prototyping and visual configuration versus Visual Studio Code's code-first workflows, version control, and local development.

Learn to build your first HR helpdesk agent in Microsoft Foundry Portal, configure GPT 4.1, system prompt, tools, knowledge, memory, and parameters, then test and save.

Learn to set up Visual Studio Code for agent development by installing the Microsoft Foundry extension and Foundry toolkit, authenticate with Azure, deploy GPT 5.1, and test agents.

Configure and manage a health care agent in VS Code with the Foundry extension, set a health care system prompt and GPT 5.1, then test it in the Foundry playground.

Extend AI agents with built-in tools to perform actions, retrieve data, and interact with external systems, using Foundry's tool catalog and the AI toolkit.

Test AI agents across happy path, edge cases, boundary tests, and multi-turn conversations with tools invocation testing, then publish via the Foundry portal or VS Code to an API endpoint.

Explore how custom tools empower ai agents to automate tasks and fetch precise data from systems. Tailor workflows to your business needs and apply customer support and inventory use cases.

Learn how Microsoft Foundry extends ai agents with custom tools by using custom functions, Azure Functions, and Logic Apps to connect to external APIs and automate workflows.

Integrate custom tools into an AI agent with the Microsoft Foundry SDK, wiring function calling, azure function, and open API tools to fetch weather details for a city.

Use the MCP, the model context protocol, to standardize tool integration for AI agents, creating a consistent bridge to external platforms like GitHub, Slack, and HubSpot.

Map the MCP architecture in a practical AI app, where the host handles prompts, the client manages MCP communication, and servers provide databases, APIs, and automations.

Discover how MCP enables dynamic tool discovery for AI agents, allowing automatic discovery of tools from an MCP server, generating wrappers and using tools across different llms with standardized security.

Connect the AI agent to remote MCP servers, automatically discover tools, wrap them in async functions, and assemble a dynamic function tools set for use.

Demonstrates MCP server integration with an AI agent using Microsoft Learn and Foundry SDK; configure the OpenAI client and MCP tool, then run a conversation to generate Azure CLI commands.

Learn how retrieval augmented generation (RAG) enables AI agents to access real-time enterprise knowledge, mitigate hallucinations, and deliver grounded, transparent responses by retrieving, augmenting, and generating from enterprise sources.

Learn how Microsoft Foundry IQ delivers a centralized knowledge platform that unifies RAG pipelines and a shared knowledge layer for AI agents, built on Azure AI Search.

Configure data sources behind the knowledge base using Azure AI Search Index, Azure Blob Storage, web connectors, SharePoint remote or SharePoint Indexed, One Lake, and Microsoft Fabric.

Learn how to retrieve accurate, cited information from Microsoft Foundry IQ, grounding AI answers in trusted enterprise data and referencing original documents.