IBM engineer Tejas Kumar launches 3-week live course on AI agent evals and harnesses
TejasKumar_ · x · 2026-11-20
IBM AI engineer and O'Reilly author Tejas Kumar is running a three-week live Maven course (Jan 13-27, $1,999) on making AI agents reliable, targeting the gap where agents work in demos but quietly fail in production with nobody able to state the failure rate.
- Week 1: read real traces, name failure modes, turn them into code checks and an LLM judge validated against your own labels
- Week 2: build a harness from an empty loop — tools, context management, guardrails, retries and a verify step so agents fail honestly
- Week 3: live debugging on participants' own agents
Graduates leave with evals and a harness that work on any model.
More from coding & agent
- A 2026 company is just a folder: every department as .md agent files — mdancho84 · 2026-11-20
- Google AI Studio can now build Android apps from a prompt and install them privately, no Play Store listing needed — TawohAwa · 2026-11-20
- No task tracker fits agentic dev: 10-100 concurrent tasks need new patterns — bigblueboo · 2026-11-20
- SelfGraph: open-source versioned knowledge graphs for AI agents over any graph DB — JeremyCMorgan · 2026-11-20
- Why one researcher's immunology model was the code he'd only ever write by hand — iskander · 2026-11-20
- SoduWork: open-source agent workbench adds AI review + human approval to verify agent output — soloboy4567 · 2026-11-20