Learn LLM Path / pillar 9 of 10
Deeper track: heavier resources for job-ready depth. Concepts you already checked stay checked.
stop guess-and-tweak. Evaluation-driven development is the single biggest predictor of agent-building success (per Andrew Ng)
- Sign in to track this item
Objective vs LLM-judge; error analysis; traces KEY
Agentic AI - Module 4 (Andrew Ng) ↗videoDEEPin course
- Sign in to track this item
RAG evals (RAGAS): faithfulness, precision, recall
RAG Evaluation, RAGAS explained - Logical Lenses ↗videoDEEP12m
- Sign in to track this item
RAGAS in code (run in CI)
RAGAS Evaluation Tutorial (local, no API) - write-up ↗articleDEEPread
- Sign in to track this item
Tracing/observability in practice
LangSmith docs ↗docsDEEPdocs
Deeper · Extras
- Sign in to track this item
Serving an LLM app behind an API
FastAPI + async streaming responses (SSE/WebSockets)videoreuse your Node/React skills for the frontend⚠ link unverified
- Sign in to track this item
Containerizing AI apps
Docker image with model deps, vector DB sidecararticleyou already know this - apply it⚠ link unverified
- Sign in to track this item
Vector DB hosting choices
Pinecone/Weaviate/pgvector/Qdrant - managed vs self-hostedvideopick one, go deep⚠ link unverified
- Sign in to track this item
Cost & latency optimization
caching, prompt compression, batching, streaming-first UXvideointerview favorite topic⚠ link unverified
- Sign in to track this item
Rate limiting & retries against provider APIs
exponential backoff, fallback modelsvideoproduction reliability⚠ link unverified
- Sign in to track this item
CI/CD for AI apps
eval suite (RAGAS) gating deploys, not just unit testsvideoties back to Phase 10⚠ link unverified
- Sign in to track this item
Monitoring in production
token usage, latency, error rates, drift in eval scoresvideoLangSmith / custom dashboards⚠ link unverified