RAG patterns we actually use
A field guide to retrieval-augmented generation in production — chunking, recall, evals, and the small things that quietly matter.
RAG patterns we actually use
Retrieval-augmented generation is having its moment. Here's what we've actually shipped — and what we'd skip next time.
Chunking is half the battle
Pick a strategy that respects your document structure. Sliding windows lose semantics; section-aware chunking keeps them.
Hybrid retrieval beats pure vector
For most enterprise corpora, a BM25 + vector ensemble outperforms either approach alone.
Evals from day one
Build a small set of canonical queries with expected answers. Run them on every prompt change. You'll thank yourself in week three.
Get next week's edition in your inbox
Field notes on AI, analytics, and shipping. Weekly. Free.
SubscribeKeep reading
All posts →RAG Evals: Proven Patterns That Deliver Real Results
Explore effective retrieval-augmented generation (RAG) evaluation patterns that drive reliable AI outcomes. Learn practical strategies for robust RAG implementations.
Introduction to AI and Its Key Types: A Practical Guide for Data Professionals
Explore the foundational types of artificial intelligence and how they apply to analytics and business operations. Understand AI beyond the buzzwords.
How to Write a Data Analytics Resume in 2026
Learn how to craft a standout data analytics resume for 2026 with actionable tips and examples to enhance your job prospects in the analytics field.