How RAG Actually Works: A Practical Guide
Retrieval-augmented generation, explained the way you would build it: chunking, embeddings, vector search, and the failure modes nobody mentions.
Blog
Code-first writing on AI engineering, backend systems, and the craft of software — no fluff, no recycled documentation.
Showing all 6 articles
Retrieval-augmented generation, explained the way you would build it: chunking, embeddings, vector search, and the failure modes nobody mentions.
Twelve things to verify before your FastAPI service takes real traffic — connection pools, timeouts, validation, and the mistakes that page you at 3 a.m.
Type hints are not bureaucracy — used well, they catch real bugs, document intent, and make refactoring safe. Here is the practical subset worth learning.
Shipping an LLM feature without evaluation is shipping untested code. A practical framework: golden sets, graded rubrics, and regression gates.
Most scaling problems are read problems, and most read problems are solved by caching — if you can answer the two hard questions: where, and how stale.
You do not need another course before you build something with AI. You need a small project, a deadline, and permission to build something imperfect.