Bob van Luijt about the origin, business and use cases of the Weaviate vector search engine
Timecodes
00:00 - Intro
00:47 - Bob's background and how Weaviate came about
03:50 - Was it a need for vector search or that GloVE / word2vec have enabled you to create Weaviate?
05:27 - Unstructured data is huge: "If you know what you are looking for, you can find it. If you don’t know what you are looking for, you can't". How can vector search help here?
7:29 - How Bob set up an experiment with disambiguating "apple" with word vectors and centroids
11:53 - Will contextualized embeddings supersede XV century old inverted index data structure?
WHAT
15:05 - What is it that you are building at SeMI, where is your focus and what use cases you go after?
22:36 - The difference and connection between a Module and a Model. And why GraphQL API? How about Python?
27:31 - The overlap between the tech and business layers is expressed as an API
31:16 - Pizza orders as a use case for vector search. Bob's top-down approach to identify new use cases
34:12 - Are many products today still stuck with the Inverted Index? Can they do better?
39:18 - Electric car batteries and the Layerying problem
HOW
45:47 - Some vector databases are close source, some are open source. Why Weaviate is open source, and how does it support its business model?
1:02:30 - Open source can sparkle an idea you can try right away
1:04:15 - Why did you customize the ANN algorithm called HNSW?
1:10:35 - When to go deep into data types / data structures to optimize performance?
WHY
1:17:11 - Why do you personally work in the field of vector search?
1:20:16 - Is there a specific use case you particularly like?
1:22:36 - Weaviate and huge Wikipedia dataset -- available soon!
1:27:07 - Announcement from Bob!
Show notes:
1. Layering problem: [ Ссылка ]
2. Podcast with Etienne Dilocker (SeMI Technologies Co-Founder & CTO): [ Ссылка ]
3. SOC2: [ Ссылка ]
4. Dmitry's post on 7 Vector Databases: [ Ссылка ]
5. Billion-Scale ANN Challenge: [ Ссылка ]
6. Weaviate Introduction: [ Ссылка ]
Newsletter: [ Ссылка ]
7. Use case: Scalable Knowledge Graph Search for 60+ million academic papers with Weaviate: [ Ссылка ]
8. Bob's Twitter: [ Ссылка ]
9. Dmitry's Twitter: twitter.com/DmitryKan
10. Dmitry's tech blog: [ Ссылка ]
Audio version of this episode:
Apple Podcasts: [ Ссылка ]
SoundCloud: [ Ссылка ]
