Advanced — Agents และ Evals

Who it’s for

You build for a living and you’ve shipped at least one AI feature to production. You want to go deep on the two hardest problems: agentic loops and evaluation at scale.

Format

• 6 weeks, ~8 hours per week
• Twice-weekly live sessions
• Mandatory shipping by week 4 — the second half of the cohort is hardening the system you built
• Pair you with a lab mentor for the duration

Curriculum

1. Week 1 — Agent loops: planning, tool use, recovery
2. Week 2 — Memory and context window discipline
3. Week 3 — Multi-agent: when it helps, when it hurts
4. Week 4 — Eval pipelines: building a harness that catches regressions
5. Week 5 — Observability: tracing, drift detection, alerting
6. Week 6 — Productionization: cost, latency, safety, rollout