Generative AI & Agentic Solutions
This domain is the heaviest section of the AI-103 exam (35-40%). It covers building generative AI applications with Azure OpenAI and Foundry, implementing RAG pipelines, designing agentic architectures, and orchestrating multi-step AI workflows.
AI agents go beyond simple prompt-response patterns. They combine LLM reasoning with tool use (function calling), memory, and multi-step planning. You'll learn to build both generative AI apps and autonomous agents that can query databases, call APIs, process documents, and coordinate with other agents to solve complex problems.
What You'll Learn
- Deploy and manage models with Azure AI Foundry
- Implement RAG (Retrieval-Augmented Generation) patterns
- Design prompt engineering strategies for production
- Build multi-model orchestration flows
- Design agent architectures with planning, memory, and tools
- Implement function calling and tool use patterns
- Build multi-agent orchestration and collaboration
- Handle agent safety boundaries and error handling
Skills Measured
- Create and configure Azure AI Foundry projects
- Deploy and manage Azure OpenAI models
- Implement RAG pipelines with Azure AI Search
- Apply prompt engineering techniques
- Configure content filtering and safety
- Design agent architectures for complex scenarios
- Implement tool use and function calling
- Implement multi-agent orchestration patterns
Challenges
| # | Title | Key Topics |
|---|---|---|
| 11 | Azure AI Foundry Setup | Hub/project model, connections, resource hierarchy |
| 12 | Deploy Generative AI Models | Model selection, deployment configurations |
| 13 | Prompt Flow | Orchestration, flows, evaluation |
| 14 | RAG Pattern: Basic | Chunking, embeddings, vector search, grounding |
| 15 | RAG Pattern: Advanced | Hybrid search, re-ranking, citation handling |
| 16 | Azure OpenAI Configuration | TPM, RPM, quotas, deployment options |
| 17 | Prompt Engineering | System prompts, few-shot, chain-of-thought |
| 18 | Content Filtering | Severity levels, custom filters, blocklists |
| 19 | Code & Image Generation | DALL-E, code generation, function calling |
| 20 | Multi-Model Orchestration | LangChain, Semantic Kernel, routing |
| 21 | Agent Architecture & Design | Agent components, planning loop, memory, tools |
| 22 | Tool Use & Function Calling | Tool definitions, parallel calls, response handling |
| 23 | Multi-Agent Orchestration | Agent handoffs, collaboration, Azure AI Agent Service |
Prerequisites
- Completed Domain 1 (Plan & Manage) or equivalent knowledge
- Azure OpenAI resource provisioned
- Python or C# programming skills
- Understanding of REST APIs and async patterns