Become a savvy vector database user and buyer. Come hang with me and Adam Hevenor as we learn about vector retrieval from first principles the week of August 10th.
How to choose a vector database (Aug 10) #
Sign up: https://maven.com/p/59a7d9/how-to-choose-a-vector-database I am asked constantly which vector database to choose. It’s not a simple question to answer. And I don’t just want to rattle off what I’m personally comfortable with.
Adam’s going to give us a more systematic approach to vector database selection. What should you measure? What features matter? What doesn’t get attention in vector benchmarks and comparisons?
Why your hybrid search sucks (Aug 12) #
Sign up: https://maven.com/p/ec3299/why-your-hybrid-search-sucks A classic search team mistake: naively merging lexical and dense vector search and saying “done”. User intent and query understanding must come first. Then you can consider how to select candidates for later ranking. In this talk, we’ll go beyond naive RRF, and build hybrid search that ranks and filters with the user first, not the technology.
Build your own vector database (Aug 14) #
Sign up: https://maven.com/softwaredoug/vectordb I don’t recommend you work in search without becoming intimately acquainted with vector search primitives.
- Quantization and dimensionality reduction - how to shrink the footprint of embeddings so that vector storage doesn’t cost a fortune
- Graph and clustering algorithms - how do production vector databases find nearest neighbors? How, exactly, do the core family of algorithms work?
- Filtering by metadata - Almost every production vector search will have filters. Filtering zeros in on the best candidates worth ranking. But it’s also a blind spot for production teams.
So join me in a short, paid course Friday afternoon where we’ll build a vector database together, making you an informed buyer and user.
Upcoming course: Build your own vector database
Want to understand what makes embedding retrieval fast, relevant, and useful in real AI systems? Join Build your own vector database and build the core pieces yourself, from embeddings and indexing to search and retrieval.