AI Instructor Live Labs Included Exam Prep

AI-901: Microsoft Azure AI Fundamentals for Python Developers

Foundational, hands-on AI-901 prep. Learn Responsible AI, Foundry models, prompts, agents, text, speech, vision, and Content Understanding — all on Microsoft Foundry using Python.

Beginner

1d 3h 35m

22 Lessons

AI-901

View badge details

Exam Preparation Included

Practice with real exam-style questions for the AI-901 certification. AI-powered feedback helps you understand every answer.

Practice Questions Track Progress AI Feedback

About This Course

Get hands-on with Microsoft Azure AI and Microsoft Foundry to prepare for the AI-901 certification exam. This foundational, AI-led course walks you through the two exam objectives: identifying AI concepts and capabilities (responsible AI, model components, AI workloads) and implementing AI solutions using Microsoft Foundry (generative AI apps & agents, text & speech, computer vision & image generation, and information extraction with Azure Content Understanding). Every topic is paired with a guided hands-on lab in the browser-based Python + AI VS Code environment. You will build chat clients, agents, text analysis apps, speech-enabled applications, vision/image-gen experiences, and a Content Understanding document pipeline all against a real Azure subscription before finishing with a capstone multimodal Foundry assistant.

Course Curriculum

22 Lessons

AI Lesson

Responsible AI Principles

45m

Learn Microsoft's six principles of Responsible AI — fairness, reliability & safety, privacy & security, inclusiveness, transparency, and accountability — and how Azure AI Content Safety enforces them inside Microsoft Foundry. Taught by Skill Me UP Steve.

Lab Exercise

Responsible AI Principles — Lab Exercises

1h 40m 3 Exercises

Hands-on with Azure AI Content Safety in Microsoft Foundry — run harm-category evaluations, jailbreak/prompt-shield detection, and map results to the six Responsible AI principles.

Provision Azure AI Content Safety Create an Azure AI Content Safety resource in the pre-created resource group, capture its endpoint, and verify that your lab identity has the Cognitive Services User role needed for keyless access. ~15 min

Run the Content Safety Classifier Execute src/main.py against the five sample prompts in prompts.json and interpret the Hate, Violence, Sexual, and SelfHarm severity scores returned by the Content Safety API. ~20 min

Map Outputs to the 6 Responsible AI Principles Add two edge-case prompts to prompts.json, rerun the classifier, then write a short reflection document mapping observed severity scores to Microsoft's six Responsible AI principles. ~20 min

AI Lesson

AI Models & Configurations

45m

Understand how generative AI models work under the hood, how to pick a model in the Foundry catalog, and how to tune deployment configuration parameters (temperature, top_p, max tokens, system prompt, etc.).

Lab Exercise

AI Models & Configurations — Lab Exercises

1h 45m 3 Exercises

Browse the Foundry model catalog, create a Foundry project, deploy a chat model, and test it from the playground and the Foundry SDK — tuning temperature, top_p, and max tokens to see how outputs change.

Create a Foundry Hub and Project Sign in to Azure CLI, create an Azure AI Foundry hub and project in the pre-created resource group, and capture the project endpoint URL needed by the Python scripts. ~20 min

Deploy a Chat Model and Inspect It Deploy gpt-4o-mini to the Foundry project via the portal and then replicate the deployment in code using src/deploy_model.py; list all deployments with src/list_models.py and explore token-per-minute quota. ~20 min

Compare Model Families and Document Your Findings Deploy a second model family (text-embedding-3-small) alongside gpt-4o-mini, list both, and write a comparison document contrasting generative chat models against embedding models in terms of task fit, pricing, and latency. ~20 min

AI Lesson

AI Workloads Overview

1h 0m

Tour the five AI workload families on Azure — generative & agentic AI, text analysis, speech, computer vision, and information extraction — and learn how to pick the right Azure service for a real-world scenario.

Lab Exercise

AI Workloads Overview — Lab Exercises

1h 40m 3 Exercises

Scenario-matching workshop — read real business descriptions, identify the right AI workload family, pick the Azure service, and validate your choice by calling that service from Python.

Set Up the Workload Identifier CLI Sign in to Azure CLI, configure a Foundry chat deployment endpoint in .env, and run src/identify.py against the ten pre-loaded scenarios to verify the classifier is working. ~15 min

Classify Real-World Scenarios and Correct Misclassifications Examine all ten classifier outputs, evaluate the accuracy of each WORKLOAD decision against your own knowledge, and document any corrections or boundary-case disagreements in docs/corrections.md. ~20 min

Add Your Own Scenarios and Reflect on Service Fit Append three original scenarios drawn from your own industry or experience to scenarios.txt, rerun the identifier, and write a brief reflection on which Azure service would best address each new scenario. ~20 min

AI Lesson

Prompt Engineering & Portal Chat

45m

Craft effective system and user prompts for generative AI models, then deploy a model and interact with it in the Foundry portal playground.

Lab Exercise

Prompt Engineering & Portal Chat — Lab Exercises

1h 45m 3 Exercises

Use the Foundry playground to iterate on system prompts, tune parameters, and test prompt-engineering patterns (role, few-shot, chain-of-thought, JSON-only output) against a deployed chat model.

Run and Compare Prompt Styles Sign in, configure a Foundry endpoint, and use src/run_prompt.py to send the same user message through each of the three pre-built system prompts — helpful-assistant, strict-classifier, and json-responder — observing how the prompt style shapes model output. ~20 min

Modify Prompts and Observe Quality Changes Edit the three system prompt files to create degraded and improved variants, rerun src/run_prompt.py for each variant, and observe how small wording changes in a system prompt can meaningfully improve or break output quality and format compliance. ~20 min

Evaluate, Document Findings, and Pick a Winner Run src/eval.py across all three original prompt styles, document pass/fail results and output quality observations in docs/evaluation.md, and write a justified recommendation for which prompt style best fits a production API integration scenario. ~20 min

AI Lesson

Foundry SDK — Building a Chat Client

45m

Move from portal to code — build a lightweight Python chat client against your Foundry deployment using the azure-ai-inference SDK and DefaultAzureCredential, then add streaming and message history.

Lab Exercise

Foundry SDK — Building a Chat Client — Lab Exercises

1h 45m 3 Exercises

Take the ai-901-chat-client starter project and turn it into a full REPL chat app — add a system prompt, streaming responses, a conversation history buffer, and keyless auth with DefaultAzureCredential.

Wire Keyless Auth and Send a First Message Sign in with az login, configure FOUNDRY_PROJECT_ENDPOINT in .env, run src/chat.py with a test message, and trace how DefaultAzureCredential obtains a bearer token to authenticate the ChatCompletionsClient call. ~15 min

Build a Multi-Turn Conversation Loop Extend src/chat.py to maintain a message history list across turns, send a 3-turn conversation that demonstrates context retention, and commit the changes to the local git repository. ~25 min

Tune System Prompts and Parameters Add --system and --temperature command-line arguments to src/chat.py, run two variants (retail concierge at temperature 0.2 and a creative persona at temperature 1.0), compare tone and consistency, then document findings in docs/tuning.md. ~20 min

AI Lesson

Single-Agent Solutions in Foundry

45m

Learn how Foundry agents differ from raw chat completions — agents, threads, runs, tools — and design a single-agent solution that uses a custom Python function tool.

Lab Exercise

Single-Agent Solutions in Foundry — Lab Exercises

1h 55m 3 Exercises

Build a single-agent solution in the Foundry portal, then connect a lightweight Python client to it — adding a custom function tool, sending messages, and interpreting the run results.

Create a Foundry Agent with a Custom Tool Provision an Azure AI Foundry agent backed by gpt-4o-mini, attach the get_store_hours Python function tool, and capture the resulting AGENT_ID for use in subsequent exercises. ~20 min

Invoke the Agent with a Tool-Using Prompt Implement the chat_once stub in src/agent.py, send a store-hours question, then trace the thread → run → tool call → tool output → final message lifecycle. ~25 min

Add a Second Tool and Test Branching Add a check_inventory(sku) function tool to agent.py, update the agent with both tools, then send prompts that exercise each code path. Cleanup the agent and threads at the end. ~25 min

AI Lesson

Text Analysis with Foundry

40m

Learn the common text analysis techniques tested on the AI-901 — keyword extraction, entity recognition, sentiment analysis, and summarization — and see how Azure AI Language and Foundry-hosted LLMs compare for each.

Lab Exercise

Text Analysis with Foundry — Lab Exercises

1h 50m 3 Exercises

Build a lightweight text analysis application against a small corpus of product reviews using the azure-ai-textanalytics SDK — extract key phrases, recognize named entities, score sentiment, and summarize.

Provision a Language Service Resource Create a Cognitive Services account of kind TextAnalytics in eastus2, capture the endpoint, and grant the Cognitive Services User role to your lab identity. ~15 min

Implement Language Detection and Sentiment Analysis Fill in the detect_language() and analyze_sentiment() stubs in src/analyze.py using TextAnalyticsClient, run the script against reviews.txt, and interpret the confidence scores. ~25 min

Implement Key Phrase Extraction and Entity Recognition Fill in the extract_key_phrases() and recognize_entities() stubs, rerun analyze.py against all five reviews, and document findings in docs/insights.md. Cleanup the Language resource at the end. ~25 min

AI Lesson

Speech AI with Foundry Tools

45m

Cover the features of speech recognition and speech synthesis, the difference between Azure Speech and multimodal chat models that accept audio, and when to use each.

Lab Exercise

Speech AI with Foundry Tools — Lab Exercises

1h 40m 3 Exercises

Build a speech-enabled app — transcribe audio with Azure Speech, synthesize voice responses, and then wire a spoken prompt straight into a multimodal model via Foundry.

Provision Azure Speech Service Create a Cognitive Services account of kind SpeechServices in eastus2, retrieve KEY1 and the region, and store them in .env for use by the Speech SDK. ~15 min

Transcribe a Short WAV File Obtain or record a short WAV file, place it in sample_data/, and run src/speech_to_text.py to transcribe it with Azure Speech STT. Examine the transcript and recognition reason code. ~20 min

Synthesize Speech with a Neural Voice Run src/text_to_speech.py to synthesize audio using en-US-AvaMultilingualNeural, iterate on a second voice and language, and document your findings. Cleanup the Speech resource at the end. ~20 min

AI Lesson

Computer Vision & Image Generation

45m

Learn the core computer-vision capabilities on Azure — image analysis, OCR, object detection, multimodal vision-in-chat — plus image generation models available through Foundry.

Lab Exercise

Computer Vision & Image Generation — Lab Exercises

1h 50m 3 Exercises

Interpret retail-shelf images with a multimodal model, call azure-ai-vision for structured analysis, and generate new marketing imagery with a Foundry image-generation model.

Provision Azure AI Vision and Confirm a DALL-E Deployment Create a Cognitive Services account of kind ComputerVision in eastus2, capture the endpoint, grant Cognitive Services User, and confirm a dall-e-3 deployment exists in your Foundry project. ~20 min

Analyze an Image Run src/analyze_image.py against sample_data/retail_shelf.jpg to extract captions, tags, detected objects, and OCR text. Then supply a second student-chosen image and compare results. ~20 min

Generate a Product Mockup with DALL-E Run src/generate_image.py to generate a marketing poster with DALL-E 3 via Foundry, iterate on two prompt variants, save the best to docs/, and clean up Vision resources. ~25 min

AI Lesson

Information Extraction with Content Understanding

50m

Extract structured information from documents, images, audio, and video using Azure Content Understanding in Microsoft Foundry Tools — analyzers, schemas, and output shapes.

Lab Exercise

Information Extraction with Content Understanding — Lab Exercises

1h 30m 4 Exercises

Build a lightweight information-extraction pipeline — define a Content Understanding analyzer, run it against a sample invoice, a retail image, and an audio clip, and return structured JSON.

Create a Content Understanding Analyzer Authenticate with Azure, define an invoice schema, and deploy a Content Understanding analyzer via the REST API. ~20 min

Extract Structured Data from a Document Run the document extraction script against a sample invoice, implement the polling loop, and inspect the structured JSON fields returned by the analyzer. ~20 min

Extract Data from an Image Implement the image extraction stub, upload a receipt image, run the image analyzer, and compare the results against document extraction. ~25 min

Extract Structured Data from Audio Create an audio analyzer for call-center intents, implement the audio extraction stub, run it against a WAV file, and clean up all analyzers when done. ~25 min

AI Lesson

Capstone Briefing

30m

Brief overview of the Northwind Horizon capstone scenario, the multimodal Foundry assistant students will build, and how each capability maps back to the AI-901 exam objectives.

Lab Exercise

Capstone Briefing — Lab Exercises

2h 0m 5 Exercises

Capstone — wire a FastAPI multimodal assistant for Northwind Horizon that combines Foundry chat, Azure Speech, vision-in-chat, and Content Understanding into one coherent app.

Wire /chat with the Foundry Chat Client Start the FastAPI app, implement the /chat route using the Foundry ChatCompletionsClient, and verify the endpoint responds to a user message. ~20 min

Wire /transcribe with Azure Speech Implement the /transcribe route to accept a WAV upload, run Azure Speech SpeechRecognizer, and return the transcript text. ~20 min

Wire /vision for Multimodal Chat Implement the /vision route to base64-encode an uploaded image and send it as an image_url content block to a multimodal gpt-4o-mini deployment, returning the model's answer. ~20 min

Wire /extract with Content Understanding Implement the /extract route using the Content Understanding poll-loop pattern, POST an uploaded file to the invoice analyzer, and return the structured fields JSON. ~25 min

End-to-End Demo and Reflection Exercise all four endpoints in sequence with sample assets, document the results in docs/demo.md, and reflect on how the four AI modalities compose into a single cohesive assistant. Clean up Azure resources when done. ~20 min

This course includes:

24/7 AI Instructor Support
Live Lab Environments
11 Hands-on Lessons
Exam Prep Included
Completion Badge

Earn Your Badge

Complete all lessons to unlock the AI-901: Azure AI Fundamentals — Python Developers achievement badge.

Skill Level Beginner

Total Duration 1d 3h 35m

AI-901: Microsoft Azure AI Fundamentals for Python Developers

Exam Preparation Included

About This Course

Course Curriculum

Responsible AI Principles

Responsible AI Principles — Lab Exercises

AI Models & Configurations

AI Models & Configurations — Lab Exercises

AI Workloads Overview

AI Workloads Overview — Lab Exercises

Prompt Engineering & Portal Chat

Prompt Engineering & Portal Chat — Lab Exercises

Foundry SDK — Building a Chat Client

Foundry SDK — Building a Chat Client — Lab Exercises

Single-Agent Solutions in Foundry

Single-Agent Solutions in Foundry — Lab Exercises

Text Analysis with Foundry

Text Analysis with Foundry — Lab Exercises

Speech AI with Foundry Tools

Speech AI with Foundry Tools — Lab Exercises

Computer Vision & Image Generation

Computer Vision & Image Generation — Lab Exercises

Information Extraction with Content Understanding

Information Extraction with Content Understanding — Lab Exercises

Capstone Briefing

Capstone Briefing — Lab Exercises

This course includes:

Earn Your Badge

AI-901: Azure AI Fundamentals — Python Developers

Skills You'll Earn