Box taps Nvidia’s Nemotron for new agent-first content retrieval system
The system can supposedly answer queries four times faster, with 20% fewer errors and 90% fewer tokens used.
TL;DR
- Box is incorporating Nvidia's Nemotron embedding system into its core product with the new Box Content Retrieval feature.
- This feature aims to make agents answer queries more effectively by using Nemotron-generated metadata.
- Box claims users will experience faster, more accurate, and less expensive queries.
- Nvidia's VP expects queries to be answered four times faster, with 20% fewer errors and 90% fewer tokens.
- The system will be available to Box Enterprise Advance customers in early next year.
- Box CEO Aaron Levie highlighted that the system boosts search relevance by sending relevant document chunks instead of requiring agents to read entire documents.
- The feature uses Nvidia's Nemotron Embed project, focusing on open-weight embedding models for automated retrieval systems.
- The system is expected to be particularly useful for professionals like lawyers, helping them find relevant clauses and precedents.
- Nemotron's embeddings map underlying concepts, allowing searches to surface related themes even with different wording.
- While Nemotron models are free, the system also relies on Box's existing metadata and permission controls.