
Unsiloed AI raises $7.5 million to structure enterprise documents for AI agents
The AMW Read
The seed round adds a document-processing player to a crowded data infrastructure market, while its reported accuracy remains unverified.
Unsiloed AI raises $7.5 million to structure enterprise documents for AI agents
Unsiloed AI raised $7.5 million in seed funding led by Felicis, with participation from Y Combinator, Orange Collective, Uncorrelated Ventures, and angel investors. The San Francisco startup extracts information from complex documents and shows where it came from in the source. CEO Aman Mishra said customers include a bank, a law firm, and voice AI startup Bolna AI. The company plans to use the funding for hiring, sales, and proprietary datasets.
The funding adds another specialist to a crowded data infrastructure market that includes Reducto, Unstructured, Microsoft, and Amazon. Unsiloed targets a problem for agent applications: tables, charts, and complex layouts can be difficult to convert into reliable data. Its approach processes text and images in separate streams and directs more computing power to areas such as figures and charts. Source references give customers a way to check extracted information against the original document.
For builders and investors, the test is whether extraction quality holds up across a customer's own documents and workflows. Unsiloed reported 99.78% accuracy across roughly 84,000 data points from 542 financial documents, but the result has not been independently verified. The company also reported its own results on the public olmOCR-Bench benchmark. Pricing starts at one cent per page, according to Mishra.