Azure AI Fundamentals (AI-901)

Section 1: Lab: AI on Azure and the AI-901 Outline

• The AI-901 exam structure: AI concepts (40-45%) and implementing solutions with Microsoft Foundry (55-60%)
• Core AI workloads: generative AI, agentic AI, text, speech, computer vision and information extraction
• What Microsoft Foundry is: resources, projects, the model catalogue and Foundry Tools
• Match workloads to business problems using sample scenarios
• Lab: create a first Foundry project in your own subscription

Section 2: Workshop: Responsible AI Principles

• Fairness, reliability and safety: outputs that are fair and dependable
• Privacy and security: protecting personal and organisational data
• Inclusiveness and transparency: designing for everyone and explaining what the AI does
• Accountability: who is responsible when AI gets a decision wrong
• Workshop: assess the risks of a use case against all six principles

Section 3: Lab: How Generative AI Works

• Tokens, embeddings and transformers explained simply
• How large language models, small language models and multimodal models differ
• Choose a model by capability: text, images, audio, reasoning, speed and cost
• Read model cards and compare models in the model catalogue
• Lab: compare the output of two models on the same task

Section 4: Lab: Deploying Models and Setting Parameters

• Model deployment options in Foundry and how they affect cost and data location
• Key parameters: temperature, top P, max tokens and stop sequences
• Endpoints, keys and keyless authentication with Microsoft Entra ID
• Lab: deploy a model and experiment in the playground with different parameters

Section 5: Lab: Text, Speech, Vision and Information Extraction Workloads

• Text analysis techniques: keyword extraction, entity detection, sentiment and summarisation
• What speech recognition and speech synthesis can do
• Computer vision and text-to-image models, with copyright considerations
• Extracting information from text, images, audio and video, and where it fits
• Lab: analyse customer reviews and have a multimodal model describe a document image

Section 6: Lab: Prompts and a Chat App with the Foundry SDK

• The roles of system prompts and user prompts
• Set the role, scope, answer format and examples (few-shot) and reduce made-up answers
• Connect to a Foundry project with Python and call a deployed model
• Keep conversation history, handle errors and control token usage
• Lab: build a command-line chat app with your own system prompt

Section 7: Lab: Building an Agent in Foundry

• How an agent differs from a chat app: instructions, a model and tools
• Create and test a single agent in the Foundry portal
• Give the agent knowledge from document files
• Lab: write a Python client application that calls the agent

Section 8: Lab: Text, Speech and Vision Apps

• Build an app that analyses text, extracts entities and summarises
• Answer spoken prompts with a multimodal model and convert speech and text with Azure Speech
• Send images in prompts for the model to interpret and create new images with generative models
• Limits and content filters for audio and image work
• Lab: a spoken question assistant and an app that writes descriptions from product photos

Section 9: Lab: Information Extraction with Content Understanding

• Extract information from documents and forms such as invoices and receipts
• Extract information from images, audio and video
• Define the fields you need and receive structured output
• Lab: an app that reads invoices and saves them as JSON

Section 10: Workshop: Review and Exam Preparation

• Map every lab back to the topics in the AI-901 outline
• Try the free practice assessment and analyse the areas to review
• Delete resources and check costs in the subscription
• Workshop: plan your revision and book the exam yourself