RAG pipelines in production: what breaks after the demo
Chunking, retrieval quality, cost and evals - the four places where a RAG demo quietly dies on the way to production, and what to do about each.
Notes on full-stack & AI/LLM engineering - what actually works in production.
RSS feed ← Back to portfolioChunking, retrieval quality, cost and evals - the four places where a RAG demo quietly dies on the way to production, and what to do about each.
What a year of running AI code review on real pull requests taught me about prompts, noise, trust and where AI actually saves review time.
The sprint format I use with clients: fixed scope, weekly demo, honest estimates - and why it keeps projects predictable.