Track · 1/5 published
5 · Production (MLOps + LLMOps)
Evals that catch regressions, guardrails against prompt injection and PII, tracing that survives a 3am page, gateways, caching, cost budgets, drift. Both MLOps and LLMOps under one roof — because in a real system they aren't separate.
After this track you can take a working prototype into production and keep it there — with a straight face on the reliability, cost, and safety questions the on-call rotation is going to ask.
Phase 6
- Offline evals that catch regressionsOutline
- Prompt injection is a real vulnerabilityOutline
- Observability that survives a 3am pageOutline
- Caching and cost controlOutline
- Design an LLM gatewayPublished