Datasite CTO on ditching in-house parsing: better parsing lifted recall 22%
AI Engineer · youtube · 2026-11-04
On the AI Engineer podcast, Datasite global CTO Raj Bakhru talks with Reducto CEO Adit Abraham about running AI at a platform handling 60% of deals over $100M.
- Parsing billions of deal-document pages is "all edge cases" — the key reason Datasite stopped building its own parsing
- His framework for deciding what his 250 engineers build vs. buy
- VDR Bench, their internal eval: better parsing raised recall by 22%
- How AI coding is changing who he hires, and where models still fall short
Highly useful for teams doing enterprise RAG, document parsing selection, and eval engineering.
More from coding & agent
- NIST publishes design tenets for agent identity management with crypto-verifiable identities — AccBalanced · 2026-11-04
- MCP's most underrated use: running your social media accounts from a chat — heyshrutimishra · 2026-11-04
- LadybugDB: an embedded graph database with Cypher, columnar execution and vector indexes, 1.9k stars — JeremyCMorgan · 2026-11-04
- Dev hits multi-credential pitfall in NewMax Provider: one key per provider doesn't scale — yangyi · 2026-11-04
- WIP 'DiffusionGearvis' Project Uses DiffusionGemma to Speed Up and Privatize Computer-Use Agents — qberthet · 2026-11-04
- Musk cites unsustainable Grok Bot costs as Hermes offers one-command agent migration — Teknium · 2026-11-04