- Unsiloed AI raised a $7.5M Felicis-led seed to build a computable document layer with a team of just seven
- Dual-stream vision models process content and layout in parallel, while heat-map chunking shifts compute toward figures and numbers
- Every extracted field traces back to its page location, and the system supports air-gapped deployment on customer servers
AI CURATED HOT STORY · 11/18/2026
Unsiloed AI Raises $7.5M to Make Documents AI-Queryable
Unsiloed AI, a seven-person startup, has raised a $7.5 million seed round led by Felicis with participation from Y Combinator and others. The company aims to build a "document computability layer" for AI agents, converting enterprise documents into structured data that agents can query directly, addressing the pain point of unstructured documents being difficult for AI to utilize effectively in high-stakes domains such as finance, healthcare, and legal.
Technically, Unsiloed AI employs a dual-stream vision model that reads document content and layout in parallel, and uses heatmap-based chunking to focus extra compute on charts and numbers, improving extraction accuracy for critical information. Each extracted field is traceable back to its original page location, ensuring auditability, and the solution supports offline deployment on customers' own servers to meet data security and compliance requirements.
The funding will be used to expand the research and engineering team, scale sales, and build proprietary datasets. As AI agents accelerate toward deployment, Unsiloed AI targets document understanding as a key bottleneck. Whether its lean team and high-precision, traceable solution can stand out in the competitive document intelligence space remains to be seen.
Hourly heat curve
14 fully observed hoursThe line compares only sources observed throughout; gap hours are interpolated to keep the trend continuous.
Coverage timeline
Total 2 reports · Latest first- Rohan Paul (@rohanpaul_ai)T2·62 pts
- TechCrunch AIT2·62 pts
- Unsiloed AI raised a $7.5M seed led by Felicis with Y Combinator participating
- Its core tech converts tables, charts and complex layouts into structured, traceable data for AI agents
- The seven-person team will use funds to hire researchers and engineers and build proprietary datasets