AI-901: Azure AI Fundamentals Exam Prep [2026 NEW]

Present the course outline and learning recommendations for the AI901 exam, detailing two core capabilities: identify AI concepts and implement AI solutions with Microsoft Foundry, and include hands-on labs.

Explore machine learning, predictive AI, and generative AI, clarify AI versus ML and supervised versus unsupervised learning, and highlight Azure Predictive AI and Microsoft Foundry.

Explore how predictive AI and generative AI complement each other to solve business problems, such as predicting baggage carousel maintenance and flight timeliness.

Learn the core generative AI jargons, including tokens, system and user prompts, and the chat completions API, and how tokens drive costs in multimodal contexts.

Deploy a Microsoft Foundry resource and project within a resource group, configure endpoints and keys, and explore the Foundry portal to access model catalog, leaderboard, and agent services.

Deploy a GPT 4.1 chat completion model in Microsoft Foundry via the model catalog, configure a global standard deployment, apply content guardrails, and explore the chat playground with custom prompts.

Learn to craft precise prompts for AI agents by focusing on goal, context, expectations, and source. Narrow, well-engineered prompts boost accuracy across language generation, code, and image tasks.

Master the Microsoft Foundry SDK by creating an AI project client to connect to a Foundry resource using endpoint and credentials, and instantiate sub-clients for OpenAI, Anthropic, or Google Gemini.

Perform a hands-on lab deploying an OpenAI model via the Foundry SDK, connecting to a GPT 4.1 deployment, authenticated with Azure CLI, and running a code-first OpenAI client.

Deploy a GPT image 1 model in Microsoft Foundry and explore the model catalog in the image generation playground. Tweak settings and prompts with reference images to practice prompt engineering.

Explore ai agents and compound ai systems that extend large language models with apis and external tools to automate workflows through dynamic runtime planning.

Create your first Foundry agent in portal, select a foundation model, enable versioning, attach tools like web search and code interpreter, then review traces for tool calls.

Perform a hands-on lab to create a basic agent in Foundry Agent Service using a code-first Python notebook, environment setup, and a Batman-themed system prompt.

Learn to build a code interpreter enabled Foundry agent in code-first workflow, read a csv of products, run python code in a sandbox, and output a column chart of prices.

Learn to build a Foundry agent with the Microsoft Learn MCP server using the Foundry SDK, environment variables, and a Python notebook to enable MCP tool calling.

Explore Azure language service capabilities via a foundry resource, including language detection, key phrase extraction, sentiment analysis, named entity recognition and entity linking, plus personally identifiable information detection and redaction.

Perform sentiment analysis, language detection, keyphrase detection, named entity recognition, linked entities, and PII recognition using Azure language service in a hands-on lab with a Python notebook.

Understand how text analysis works from tokenization and normalization through n-grams, stemming, and part-of-speech tagging, and explore TF-IDF, bag of words, text rank, and vectorization for summarization with LLMs.

Explore the Azure Translator Service Playground in Microsoft Foundry, test text and document translation across English, Arabic, and French, and call the service via SDK or REST API.

Call the Azure Translator service with a code-first approach, set up environment variables, and use the Azure AI Translation SDK in Python to translate and transliterate text.

Introduces the features and capabilities of the Azure Computer Vision service (Azure AI Vision) - a Foundry tool - with caption and tag generation, object detection, and OCR.

Explore how image processing uses pixels, grayscale and RGB colors, and how 3x3 kernels modify brightness, blur, edge detection, and train a convolutional neural network for object detection.

Explore the Azure Speech Service workflows for speech-to-text and text-to-speech, including authentication, audio input options, json responses, and neural voice synthesis.

Explore how speech processing works under the hood, including speech-to-text, text-to-speech, phonemes, MFCC, Fourier transforms, acoustic and language modeling, and post-processing.

Build an audio-enabled JNI app using a multimodal GPT audio model on Azure OpenAI, deploying the model, configuring environment variables, and sending MP3 audio input via base64 encoding.

Explore Azure Speech Voice Live with neural voices and low latency for real-time, multi-modal conversations, including Gen AI models and Python code samples.

Discover how Azure AI Content Understanding processes unstructured data from invoices, PDFs, images, audio, and video, using analyzers and standard and pro modes for structured insights and compliance checks.

Explore Azure AI Content Understanding in the Microsoft Foundry portal, using the Content Understanding Studio to test document, audio, image, and video analyzers and inspect json responses for integration.

Learn to use the prebuilt Azure AI Content Understanding Read API in a code-first lab to analyze an invoice and handwritten image, with environment setup, async analysis, and markdown results.

Leverage the prebuilt layout API in Azure AI Content Understanding to preserve invoice tabular structure, extracting a markdown table with date, item code, description, quantity, and amount due.

Explore the principles of responsible AI, including fairness, inclusiveness, reliability and safety, privacy and security, transparency, and accountability, with human-in-the-loop frameworks.

Configure content guardrails and safety filters in foundry to protect agents and foundational language models from jailbreak attempts and harmful content, using configurable controls and guardrails.

Explore red teaming in Microsoft Foundry, using evaluation tools and attack strategies to test agent safety across risk categories, with synthetic data, attack inputs, and attack success rate analysis.