Goal: Go from zero to confidently building, deploying, and managing AI solutions on Azure AI Foundry.
Pace: 6–8 hours per week · All weeks are self-paced — adjust as needed.
Prerequisites (before Week 1):
- Active Azure subscription (free trial)
- Basic Python knowledge
- Familiarity with REST APIs
- VS Code with the Azure and Python extensions installed
- Azure CLI installed (install guide)
- [ ] checkboxes.| Estimated Hours: 6 | Difficulty: Beginner |
Prerequisites: Azure subscription, Azure CLI
| # | Task | Done |
|---|---|---|
| 1 | Read What is Azure AI Foundry? | - [ ] |
| 2 | Explore the AI Foundry portal at https://ai.azure.com | - [ ] |
| 3 | Understand Hubs, Projects, and Resources hierarchy | - [ ] |
| 4 | Review AI Foundry architecture | - [ ] |
| 5 | Compare AI Foundry vs standalone Azure AI Services | - [ ] |
Lab: Lab 01 — Create Your First Hub and Project
Microsoft Learn Path: Introduction to Azure AI Foundry
| Estimated Hours: 7 | Difficulty: Beginner |
Prerequisites: Week 1 completed (Hub & Project created)
| # | Task | Done |
|---|---|---|
| 1 | Browse the Model Catalog | - [ ] |
| 2 | Understand model families: OpenAI, Meta Llama, Mistral, Microsoft Phi | - [ ] |
| 3 | Learn deployment options: Serverless API, Managed Compute, Global/Standard | - [ ] |
| 4 | Deploy a GPT model via the portal | - [ ] |
| 5 | Deploy an open-source model via Serverless API | - [ ] |
| 6 | Compare model capabilities and pricing | - [ ] |
Lab: Lab 02 — Deploy and Compare Models
Microsoft Learn Path: Deploy models in Azure AI Foundry
| Estimated Hours: 7 | Difficulty: Beginner |
Prerequisites: Week 2 completed (at least one model deployed)
| # | Task | Done |
|---|---|---|
| 1 | Use the Chat Playground with your deployed model | - [ ] |
| 2 | Experiment with system prompts, temperature, top-p | - [ ] |
| 3 | Practice few-shot prompting techniques | - [ ] |
| 4 | Try chain-of-thought prompting | - [ ] |
| 5 | Use the Images and Completions playgrounds | - [ ] |
| 6 | Export playground settings to code | - [ ] |
Lab: Lab 03 — Prompt Engineering in the Playground
Microsoft Learn Path: Prompt engineering techniques
Self-assess before continuing:
If any items are unchecked, revisit the corresponding week before proceeding.
| Estimated Hours: 8 | Difficulty: Intermediate |
Prerequisites: Weeks 1–3 completed, Python environment set up
| # | Task | Done |
|---|---|---|
| 1 | Install SDKs: azure-ai-projects, azure-ai-inference |
- [ ] |
| 2 | Authenticate with DefaultAzureCredential |
- [ ] |
| 3 | Send chat completions programmatically | - [ ] |
| 4 | Handle streaming responses | - [ ] |
| 5 | Use the REST API directly with curl |
- [ ] |
| 6 | Explore the azure-ai-inference SDK reference | - [ ] |
Lab: Lab 04 — Build a Chat App with the AI SDK
Microsoft Learn Path: Develop AI apps with Azure AI Foundry SDK
| Estimated Hours: 8 | Difficulty: Intermediate |
Prerequisites: Week 4 completed, basic understanding of vector embeddings
| # | Task | Done |
|---|---|---|
| 1 | Understand the RAG pattern | - [ ] |
| 2 | Set up Azure AI Search as a data source | - [ ] |
| 3 | Create an index with vector and keyword search | - [ ] |
| 4 | Connect your data in the Playground (“Add your data”) | - [ ] |
| 5 | Implement RAG with the SDK | - [ ] |
| 6 | Compare chunking strategies | - [ ] |
Lab: Lab 05 — Build a RAG Application
Microsoft Learn Path: Implement RAG with Azure AI Foundry
| Estimated Hours: 8 | Difficulty: Intermediate |
Prerequisites: Week 5 completed (basic RAG working)
| # | Task | Done |
|---|---|---|
| 1 | Implement hybrid search (vector + keyword) | - [ ] |
| 2 | Add semantic ranking | - [ ] |
| 3 | Experiment with different embedding models | - [ ] |
| 4 | Implement reranking strategies | - [ ] |
| 5 | Optimize chunk sizes and overlap | - [ ] |
| 6 | Handle multi-turn conversations with context | - [ ] |
Lab: Lab 06 — Advanced RAG Optimization
Microsoft Learn Path: Azure AI Search vector search
Self-assess before continuing:
| Estimated Hours: 8 | Difficulty: Intermediate |
Prerequisites: Week 6 completed
| # | Task | Done |
|---|---|---|
| 1 | Understand the Azure AI Agent Service | - [ ] |
| 2 | Create an agent with function calling | - [ ] |
| 3 | Add tools: Code Interpreter, File Search, Bing grounding | - [ ] |
| 4 | Build custom function tools | - [ ] |
| 5 | Implement agent orchestration patterns | - [ ] |
| 6 | Handle agent errors and guardrails | - [ ] |
Lab: Lab 07 — Build an AI Agent with Tools
Microsoft Learn Path: Build AI agents with Azure AI Agent Service
| Estimated Hours: 7 | Difficulty: Intermediate |
Prerequisites: Week 7 completed (working agent or RAG app to evaluate)
| # | Task | Done |
|---|---|---|
| 1 | Understand evaluation concepts | - [ ] |
| 2 | Install azure-ai-evaluation SDK |
- [ ] |
| 3 | Run built-in evaluators: Groundedness, Relevance, Coherence, Fluency | - [ ] |
| 4 | Create custom evaluators | - [ ] |
| 5 | Generate evaluation datasets | - [ ] |
| 6 | Analyze evaluation results in the portal | - [ ] |
Lab: Lab 08 — Evaluate Your AI Application
Microsoft Learn Path: Evaluate generative AI apps
| Estimated Hours: 7 | Difficulty: Intermediate |
Prerequisites: Week 8 completed
| # | Task | Done |
|---|---|---|
| 1 | Review Azure AI Content Safety | - [ ] |
| 2 | Configure content filters on model deployments | - [ ] |
| 3 | Implement prompt shields and jailbreak detection | - [ ] |
| 4 | Add groundedness detection | - [ ] |
| 5 | Build a responsible AI dashboard | - [ ] |
| 6 | Evaluate for safety metrics (violence, self-harm, hate, sexual) | - [ ] |
Lab: Lab 09 — Content Safety & Responsible AI
Microsoft Learn Path: Azure AI Content Safety
Self-assess before continuing:
| Estimated Hours: 8 | Difficulty: Advanced |
Prerequisites: Weeks 1–9 completed, understanding of when fine-tuning is appropriate
| # | Task | Done |
|---|---|---|
| 1 | Understand when to fine-tune vs RAG vs prompt engineering | - [ ] |
| 2 | Prepare training data in JSONL format | - [ ] |
| 3 | Submit a fine-tuning job via the portal | - [ ] |
| 4 | Submit a fine-tuning job via the SDK | - [ ] |
| 5 | Evaluate fine-tuned model performance | - [ ] |
| 6 | Deploy and test the fine-tuned model | - [ ] |
Lab: Lab 10 — Fine-Tune a Model
Microsoft Learn Path: Fine-tune models in Azure AI Foundry
| Estimated Hours: 8 | Difficulty: Advanced |
Prerequisites: Week 10 completed, familiarity with GitHub Actions or Azure DevOps
| # | Task | Done |
|---|---|---|
| 1 | Design an MLOps pipeline for AI Foundry | - [ ] |
| 2 | Set up CI/CD with GitHub Actions | - [ ] |
| 3 | Automate evaluation in the pipeline | - [ ] |
| 4 | Implement blue/green deployment | - [ ] |
| 5 | Configure monitoring and alerts | - [ ] |
| 6 | Set up RBAC and network security (private endpoints, managed identity) | - [ ] |
Lab: Lab 11 — CI/CD Pipeline for AI Foundry
Microsoft Learn Path: MLOps for AI applications
| Estimated Hours: 8 | Difficulty: Advanced |
Prerequisites: All previous weeks completed
| # | Task | Done |
|---|---|---|
| 1 | Work with multi-modal models (GPT-4o vision, DALL-E) | - [ ] |
| 2 | Implement speech-to-text and text-to-speech | - [ ] |
| 3 | Build multi-agent orchestration patterns | - [ ] |
| 4 | Implement real-time streaming with WebSockets | - [ ] |
| 5 | Design a hub-spoke enterprise architecture | - [ ] |
| 6 | Capstone: Build an end-to-end AI application combining 3+ techniques | - [ ] |
Lab: Lab 12 — Capstone: End-to-End AI Application
Microsoft Learn Path: Build multimodal AI applications
Congratulations! Confirm you’ve achieved the following:
| Resource | Link |
|---|---|
| Azure AI Foundry Documentation | https://learn.microsoft.com/azure/ai-studio/ |
| Azure AI Foundry SDK (Python) | https://learn.microsoft.com/python/api/overview/azure/ai/ |
| Azure AI Services Pricing | https://azure.microsoft.com/pricing/details/ai-studio/ |
| Azure AI Foundry Samples (GitHub) | https://github.com/azure-samples/aistudio-python-quickstart-sample |
| Microsoft Learn AI Training | https://learn.microsoft.com/training/browse/?products=ai-services |
| Responsible AI Resources | https://www.microsoft.com/ai/responsible-ai |