Sridhar Ramaswamy, Snowflake | Snowflake Summit 2023
summary
Describes a working generative AI search built by a 50-person team that crawls hundreds of millions of pages daily, runs a 5–6 petabyte index and serves billions of pages, fine-tunes and deploys models in its own clusters rather than relying on OpenAI, and achieves ~3–4% paid conversion on consumers. Plans integrate retrieval, sentence-level references and model execution into Snowflake via Snowpark Containers, enabling conversational SQL, catalog search, model hosting from Hugging Face and secure in‑perimeter inference.
AI-written summary · about
chapters (11)
- 0:00Introduction to Snowflake Summit
- 1:03Neeva's origin and mission
- 3:33Consumer search challenges and pivot
- 6:44Neeva’s fit with Snowflake
- 8:35Neeva’s tech and Snowflake’s appeal
- 10:25Snowflake’s evolution beyond SQL
- 13:39Snowpark Containers and AI integration
- 16:30Language model costs and democratization
- 19:01Future of enterprise AI with Snowflake
- 20:24Immediate enterprise applications
- 23:16Closing remarks and next segment
