Vector Databases
How vector databases turn meaning into math and find it fast — embeddings, similarity, ANN indexes, quantization, hybrid search, and running them in production.
Vector Databases: Zero to Industry Expert
A complete, plain-English course. No prior machine-learning knowledge assumed. By the end you’ll understand how vector databases work under the hood and be able to design, build, and run them in production like a senior engineer.
How to use this course
Read the modules in order. Each one builds on the last. Every module has:
- The big idea — explained like you’re five.
- The real mechanics — what’s actually happening, with the math kept gentle.
- Hands-on — code you can run.
- Gotchas — the mistakes that bite people in production.
- Check yourself — questions to make sure it stuck.
The modules
| # | File | What you’ll learn |
|---|---|---|
| 1 | 01-what-is-a-vector-database.md | What these things are and why the world suddenly needs them |
| 2 | 02-embeddings.md | How text, images, and audio become numbers (the heart of everything) |
| 3 | 03-vectors-and-similarity-metrics.md | Vector math made painless; cosine, dot, Euclidean |
| 4 | 04-similarity-search.md | Exact nearest-neighbor search and why it doesn’t scale |
| 5 | 05-ann-indexes.md | The clever tricks (HNSW, IVF, PQ, LSH) that make search fast |
| 6 | 06-quantization-and-compression.md | Shrinking vectors so billions fit in RAM |
| 7 | 07-choosing-a-vector-database.md | pgvector, Pinecone, Qdrant, Milvus, Weaviate, Chroma, FAISS |
| 8 | 08-building-rag.md | Build a real Retrieval-Augmented Generation app end to end |
| 9 | 09-hybrid-search-and-filtering.md | Combine keywords + vectors + metadata filters |
| 10 | 10-production-and-scaling.md | Sharding, replication, latency, cost, monitoring |
| 11 | 11-evaluation-and-quality.md | Measure and improve retrieval quality (recall, MRR, nDCG) |
| 12 | 12-advanced-and-frontier.md | Reranking, multi-vector, late interaction, GraphRAG, security |
| 13 | 13-glossary-and-cheatsheet.md | Every term defined + a one-page reference |
The 30-second summary (so the whole course has a spine)
A vector database stores data as lists of numbers called embeddings. Embeddings are created by AI models so that things with similar meaning end up close together in number-space. To “search,” you turn your query into an embedding too, then find the stored vectors nearest to it. Doing this exactly is slow at scale, so we use Approximate Nearest Neighbor (ANN) indexes that trade a tiny bit of accuracy for enormous speed. Everything else in this course is detail on top of that one idea.
Start with 01-what-is-a-vector-database.md.