AI Agents: Build and Evaluate

Agents with tools and multi-agent systems, built for reliability and tested like real software.

Design agents that call tools, split work across several agents, and recover from errors. Every system you build comes with an evaluation suite.

What you will be able to do

• Build agents that use tools safely
• Design router, worker, and validator agents
• Write evaluations that catch regressions

Before you start

• Build LLM Apps with RAG or similar
• Comfort with Python

Course outline

1. The agent loop: An agent with tools, a step limit, and confirmations. Your own small harness.
2. Workflows and multi-agent systems: Router, worker, and validator agents. A hybrid workflow with validation gates.
3. See inside the agent: Traces for every model and tool call. A trace you can read and debug.
4. Evaluate and harden: An eval suite that catches regressions. A release report with quality, cost, and failures.

All AI and machine learning courses