Generative AI Solutions
This domain covers building applications with Azure OpenAI and the Azure AI Foundry platform. It represents 15–20% of the AI-102 exam and has been increasing in weight as Microsoft accelerates Generative AI adoption.
You'll learn to deploy large language models, implement RAG (Retrieval-Augmented Generation) architectures that ground model responses in your own data, engineer effective prompts, configure content filtering for safety, and orchestrate complex multi-step AI workflows using tools like Prompt Flow and Semantic Kernel.
This is the most rapidly evolving domain on the exam. Focus on the patterns (RAG, prompt engineering, content filtering) rather than memorizing specific model names, as the model landscape changes quarterly.
What You'll Learn
- Create and configure Azure AI Foundry projects
- Deploy and manage Azure OpenAI model deployments
- Implement RAG (Retrieval-Augmented Generation) with your own data
- Engineer effective prompts (system messages, few-shot, chain-of-thought)
- Configure content filtering and safety mechanisms
- Generate images with DALL-E and code with GPT models
- Orchestrate AI workflows with Prompt Flow
- Manage token limits, rate limiting, and quotas
Skills Measured
- Create an Azure AI Foundry project and manage connections
- Select and deploy Azure OpenAI models
- Implement Retrieval-Augmented Generation (RAG)
- Design and optimize prompts for different scenarios
- Configure and customize content filtering
- Implement Azure OpenAI on your data
- Manage token usage, quotas, and rate limits
- Evaluate and monitor generative AI response quality
Challenges
| # | Title | Key Topics |
|---|---|---|
| 11 | Azure AI Foundry Project Setup | AI Foundry portal, projects, connections, hubs |
| 12 | Deploy Azure OpenAI Models | GPT-4o, embedding models, TPM/RPM configuration |
| 13 | Model Management & Versioning | Deployment strategies, model updates, fallback |
| 14 | RAG: Chunking & Embedding | Document chunking strategies, embedding generation |
| 15 | RAG: Vector Search & Grounding | Vector store integration, citation, grounding |
| 16 | Prompt Engineering Fundamentals | System prompts, temperature, few-shot examples |
| 17 | Advanced Prompt Techniques | Chain-of-thought, structured output, evaluation |
| 18 | Content Filtering & Safety | Severity levels, custom filters, blocklists |
| 19 | Image & Code Generation | DALL-E, function calling, structured output |
| 20 | Orchestration with Prompt Flow | Flow design, evaluation, deployment |
Prerequisites
- Completed Domain 1 (Plan & Manage) or equivalent knowledge
- Azure OpenAI access approved (apply here)
- Understanding of REST APIs and JSON
- Basic understanding of ML concepts (training, inference, embeddings)