AI Agents for Beginners

Topics covered

• •Set up the environment needed for the course (GitHub Models or Azure/Microsoft Foundry).
• •Define what an AI agent is and distinguish it from a "regular" LLM application.
• •Break down the types of agents and the three basic components of an agentic system.

Topics covered

• •Understand what agentic frameworks provide and how to choose the right one.
• •Learn the design principles of human-centered agents.
• •Master the key concepts of the Microsoft Agent Framework (MAF): agents, threads, workflows, middleware, orchestration.

Topics covered

• •Master the Tool Use pattern: function schemas, calling, and handling results.
• •Understand how agentic RAG differs from classic retrieve-then-read, and when it's worth the added complexity.
• •Learn to build a planning agent: goal decomposition, structured output, and iterative replanning.

Topics covered

• •Recognize scenarios where a multi-agent system is justified versus when a single agent is enough.
• •Know the building blocks of the multi-agent pattern: communication, coordination, architecture, observability.
• •Understand specific patterns: group chat, task hand-off, and collaborative filtering.

Topics covered

• •Understand metacognition ("thinking about thinking") and learn to build self-correcting agents.
• •Master reliability and safety techniques: the system message framework, threat modeling, and human-in-the-loop.
• •Learn to take agents to production: traces and spans, metrics, offline and online evaluation, and cost management.

Topics covered

• •Understand the three agentic protocols: MCP (agent ↔ tools), A2A (agent ↔ agent), and NLWeb (agent ↔ websites).
• •Master context engineering: types of context, management strategies, and common context failure modes.
• •Understand the types of agent memory (working, short-term, long-term, and specialized) and the tools used to store it (Mem0, Cognee, Azure AI Search).