How RAG Actually Finds Answers (Part 2): HNSW, IVF, BM25, Hybrid Search and Re-Ranking

How RAG Actually Finds Answers (Part 2): HNSW, IVF, BM25, Hybrid Search and Re-Ranking | M011 | Mehul Ligade 🔴 Part 2 of a 3-Part RAG Series In Part 1, we built the mental model. We saw how PDFs become chunks, how chunks become embeddings, and how embeddings get d inside a vector database so an AI system can search them later. That was the foundation. Now we go one layer deeper and ask the real question: o nce the knowledge is already d, how does the system actually find the right answer quickly and reliably? That is what this article is about.