← All apps
Token Factory
Cross-Source Join Advisor
Automatically test every join across data sources and flag the risky ones before anyone writes a query.
About this project
Reads dataset schemas from DataHub and uses Nebius (DeepSeek-R1-0528) to identify join candidates by reasoning over column naming and data types, then classifies each join as SAFE/RISKY/UNSAFE with Llama-3.3-70B and writes the documented connections back into DataHub. Built at the DataHub x Nebius hackathon at Entrepreneurs First SF, April 2026.
Technologies
agents
hackathon
data-quality
datahub
datahub-nebius