Unsiloed AI
3 steps1 pieces6 Sources
Unsiloed AI, a San Francisco-based startup founded by IIT Kharagpur alumni Aman Mishra and Adnan Abbas, has raised $7.5 million in seed funding led by Felicis to transform complex enterprise documents into structured data usable by AI agents. The company addresses the challenge that traditional OCR and general-purpose LLMs struggle with tables, charts, and layout-dependent meaning in financial, legal, and healthcare documents. Unsiloed uses "dual-stream" vision models and heat map-based chunking to preserve document structure while extracting information, claiming 99.78% accuracy on a customer-supplied test of 542 financial documents (unverified) and an 88.0 score on the public olmOCR-Bench benchmark, ranking first among 19 systems. The round also included participation from Y Combinator, Orange Collective, Uncorrelated Ventures, and angel investors Aarthi Ramamurthy and Beth Turner. Unsiloed plans to use the funds to hire researchers and engineers, expand sales, and develop proprietary datasets, currently operating with seven employees and serving customers including voice AI startup Bolna AI.
How it got here
Fall 2025: Unsiloed AI joins Y Combinator's Fall 2025 batch after developing its core technology.
techcrunch.com ↗November 18, 2026: TechCrunch reports Unsiloed AI raises $7.5M seed led by Felicis to make enterprise documents usable by AI agents.
techcrunch.com ↗