Intermediate
14.5h · 62 lessons
Building Production AI Agents
From prototype to reliable, observable agents in production.
4.9rating12,480 enrolledLifetime access
Start free preview
About this course
Design, build and ship AI agents that plan, use tools, recover from failure and stay observable. Includes evaluation harnesses and cost controls.
What you'll learn
- Architect agents that stay reliable under real user traffic
- Design tool interfaces that don't confuse the model
- Implement retries, timeouts, and graceful degradation
- Roll out safely with feature flags and shadow traffic
Requirements
- Comfortable shipping a backend service to production
- Basic experience calling LLM APIs
- Familiarity with Docker or a serverless platform
Curriculum
6 modules · 24 lessons
Module 1What breaks in production4 lessons
Module 1
What breaks in production
- 01.01Failure modes of naive agents
- 01.02Latency budgets
- 01.03The cost of context
- 01.04Case study: a 3am pager
Module quiz · 3 questions
Module 2Designing the tool surface4 lessons
Module 2
Designing the tool surface
- 02.01Naming and descriptions
- 02.02Argument schemas
- 02.03Idempotent tools
- 02.04Error messages the model can act on
Module quiz · 3 questions
Module 3Control flow4 lessons
Module 3
Control flow
Module quiz · 3 questions
Module 4Reliability engineering4 lessons
Module 4
Reliability engineering
Module quiz · 3 questions
Module 5Evaluation4 lessons
Module 5
Evaluation
Module quiz · 3 questions
Module 6Operating agents4 lessons
Module 6
Operating agents
- 06.01Tracing and dashboards
- 06.02Cost attribution
- 06.03On-call playbooks
- 06.04Capstone: production agent
Module quiz · 3 questions