# Onehouse puts vector search inside Hudi and Iceberg tables instead of a separate database

> Source: <https://runtimewire.com/article/onehouse-puts-vector-search-inside-hudi-and-iceberg-tables-instead-of-a-separate>
> Published: 2026-11-18 18:14:24+00:00

**Key takeaway:** Onehouse's Lakegres stores a native vector index inside existing Apache Hudi and Apache Iceberg tables in the customer's own bucket and serves similarity queries through a Postgres endpoint, avoiding a separate vector database.

# Onehouse puts vector search inside Hudi and Iceberg tables instead of a separate database

**Founder Vinoth Chandar's Lakegres indexes embeddings in customers' own buckets and serves similarity queries through a Postgres endpoint. Onehouse listed vector search as Preview in October.**

        By [RuntimeWire Staff](https://runtimewire.com/author/runtimewire-staff)
        · Published 

Onehouse founder Vinoth Chandar built Apache Hudi at Uber because its data platform needed fresher, updatable tables at a scale ordinary databases handled badly. On November 18th, 2026, Onehouse extended that same argument to AI agents: Onehouse announced vector search in Lakegres, with a native index stored inside existing and new Apache Hudi and Apache Iceberg tables, queried through a Postgres endpoint, and kept in open formats in the customer's own bucket.…
