AI Agent Fundamentals
For software engineers who can call an LLM API and now want to build systems that act, not just answer. You will learn how agents plan, choose and execute tools, recover from errors and know when to stop, then build, test and evaluate a working ReAct agent in Python.
Course Content
What turns a language model into an agent, how the perceive–reason–act–observe loop works and when it must stop, how much autonomy to grant, and which architecture fits which problem. 3 lessons, about 45 minutes.
How agents plan: classic search (BFS, DFS, A*), goals and state that a program can check, and the ReAct loop that lets an LLM agent reason between actions. 3 lessons, about 50 minutes.
How to give an agent tools it can choose correctly, execute its actions safely with timeouts, retries and idempotency, and recover when something goes wrong. 3 lessons, about 40 minutes.
Build a ReAct agent in Python that survives real use, wire planning and tool use together through the prompt and code, and test, debug and evaluate it. 3 lessons, about 45 minutes.
A full specification, with acceptance numbers, for building and evaluating a single-agent task-automation system of your own. 1 lesson, about 15 minutes.