Advanced6 weeks · Cohort-based
Building AI Agents & Automation
Design, build and debug agents that plan and act on their own. Wire them into the systems you already use, and keep them dependable once they are there.
What you will be able to do
- Design an agent that plans, calls tools, and recovers from its own mistakes
- Write tool interfaces a model can use without hand-holding
- Measure agent reliability instead of hoping for it
- Run an agent against real systems without it doing damage
Course outline
- 01What separates an agent from a prompt: tools, memory, and a loop.
- 02Tool design — the interfaces that make an agent reliable.
- 03Planning and control flow, and knowing when to keep a human in it.
- 04Memory and context: what to keep, what to retrieve, what to forget.
- 05Evals for agents: measuring something that behaves differently every run.
- 06Production — cost, latency, failure modes, and the pager at 3am.