Links

Tags

#Vector Databases #Distributed Systems #Machine Learning #Federated Retrieval

Federated Query Routing

As scientific AI workflows increasingly rely on distributed data, retrieval is no longer confined to a single vector database, making it essential to decide which sources to search. Poor routing decisions can introduce unnecessary system cost and prevent relevant knowledge from ever being retrieved.

This project investigates the design space of query routing mechanisms for federated retrieval systems and studies how routing strategies can adaptively adjust their level of geometric information based on query complexity. Our goal is to develop principled routing approaches that improve both efficiency and retrieval quality, enabling scalable knowledge access for scientific discovery platforms.

Publications

Coming soon!

Funding and Acknowledgements

This work is funded by the US Department of Energy.

People

Ian Foster
Kyle Chard
Song Young Oh