Site icon The Data Exchange

The Future of Vector Databases and the Rise of Instant Updates

Louis Brandy on the Critical Need for Incremental Updates and Metadata Filtering.


SubscribeApple • Spotify • Stitcher • Google • AntennaPod • Podcast Addict • Amazon •  RSS.

Louis Brandy is  VP of Engineering at Rockset, the real-time search and analytics database startup formed by the creators of the popular open source project, RocksDB. We discuss his early experience building large scale infrastructure to support embeddings and vector search at Meta. He described Rockset’s current vector search offering, specifically its current features and near-term roadmap. He contends that incremental updates and metadata filtering will make vector databases more powerful and versatile. We also discuss hybrid search, retrieval-augmented Large language Models, and the future of streaming architectures in the age of embeddings.

Subscribe to the Gradient Flow Newsletter

Interview highlights – key sections from the video version:

  1. His experience with vector search and embeddings at Meta: vectors before the rise of vector databases
  2. Vector Search on Rockset
  3. Embeddings, features stores, ML infrastructure
  4. Vector Search on Rockset: features available today
  5. Initial feedback from users (Vector Search on Rockset)
  6. The importance supporting metadata filtering & incremental updates
  7. RocksDB as a storage plane
  8. Incremental updates and lessons from the world of realtime analytics
  9. The need for data management systems that can support different workloads and applications
  10. Hybrid search (keyword search + neural search)
  11. Streaming architectures in the age of embeddings
  12. Looking to the future: new neural architectures, nearest neighbor alternatives, retrieval-augmented LLMs

 

Related content:


If you enjoyed this episode, please support our work by encouraging your friends and colleagues to subscribe to our newsletter:

Exit mobile version