DataEngBytes 2023 - SYD-T2-01 - Rene Essomba
summary
Vector databases store high dimensional embeddings (images, text, audio) to give LLMs memory and reduce hallucination by retrieving source fragments rather than fine tuning. They run approximate nearest neighbor indexing, often with GPU acceleration and Kubernetes partitioning, expose CRUD APIs, and support semantic search plus metadata filtering (e.g., publication date). Choose by vector dimensionality, expected scale and latency; an internal Cevo PoC works but is not production ready.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 2:33Vector databases for LLMs
- 8:42Vector vs traditional databases
- 14:25Core Vector DB functions
- 17:41Advantages and use cases
- 22:00Choosing a Vector DB
- 27:36Q&A
related talks
41mGetting Started With Vector Databases
45mA gentle introduction to Vector Databases
32mDistributed Vector Databases - What, Why, and How
01h 23mVector Databases Demystified: Discover how they work
27mUnified AI & Relational Power: Scalable Distributed Vector Search with
33mTowards a deeper understanding of retrieval and vector databases
06mWhat makes vector databases essential for RAG and AI apps?
33mVector Databases for GenAI applications