TechAmerica
SearchSubscribe
Unsiloed AI funding
Image credit Unsiloed AI

Unsiloed AI Funding Reaches $7.5M to Improve Enterprise AI Data

Unsiloed AI raised $7.5 million in seed funding, led by Felicis, to help businesses convert complex documents into structured data for AI agents. The startup reports high extraction accuracy and plans to expand its engineering team and enterprise operations.

Unsiloed AI funding has reached $7.5 million in a seed round led by Felicis, as the San Francisco startup develops technology for processing complex enterprise documents. Y Combinator, Orange Collective and Uncorrelated Ventures also participated, alongside angel investors Aarthi Ramamurthy and Beth Turner.

The company aims to make business information easier for AI agents to interpret. Its software extracts data from PDFs, tables, charts and other document formats while preserving references to the original material.

Co-founder and CEO Aman Mishra established Unsiloed with Adnan Abbas, a former classmate at the Indian Institute of Technology Kharagpur. The founders joined Y Combinator’s Fall 2025 program after developing their initial technology.

Unsiloed AI funding supports specialized document processing

Enterprise documents frequently contain complex layouts, financial figures and industry-specific terminology. Conventional AI systems can misinterpret these elements, producing errors that affect downstream business decisions.

Unsiloed AI addresses the problem using proprietary dual-stream vision models. One stream processes text tokens, while another examines visual information from document images.

The system also uses heat map-based chunking to identify areas requiring more detailed processing. It directs additional computing resources toward complex sections, including charts, numerical values, and tables.

A specialized decoder considers industry terminology when interpreting extracted information. This helps distinguish terms that carry different meanings across financial, legal and healthcare documents.

Unsiloed also preserves references to the original documents. These allow users to inspect extracted information instead of relying entirely on AI-generated answers.

TechAmerica.ai previously examined AI audit evidence traceability, which explains why businesses need to connect extracted figures to their supporting records.

Company reports 99.78% accuracy in enterprise testing

Unsiloed reported 99.78% extraction accuracy in a test involving 542 financial documents provided by an enterprise customer. The evaluation covered approximately 84,000 data points.

The startup compared its results with major AI models, Microsoft’s Azure Document Intelligence and the customer’s existing extraction system. The reported accuracy has not been independently verified.

In a separate olmOCR-Bench evaluation, Unsiloed reported a score of 88.0. Reducto’s agentic system scored 66.0, while Unstructured received 39.9 in the company’s comparison.

The benchmark includes 1,403 PDF pages and tests mathematical expressions, tables, scanned documents and complicated layouts. The comparisons were conducted or compiled by Unsiloed and should not be treated as independent performance certification.

Unsiloed says it already serves organizations in finance, healthcare and legal services. One banking customer uses the software for document processing, while a law firm runs it on private servers.

The law firm’s deployment handles contracts and historical case files without requiring all documents to be processed through a shared cloud service.

Voice AI startup Bolna AI also uses Unsiloed to process knowledge bases supporting its conversational agents.

The technology overlaps with other enterprise automation systems. TechAmerica.ai previously reported on Dili’s AI document processing platform, which converts complex business records into structured compliance data.

Startup plans engineering expansion after funding.

Unsiloed competes with specialized document-processing companies, including Reducto and Unstructured. Microsoft, Amazon and general-purpose AI model providers also offer competing extraction capabilities.

The startup charges approximately one cent per document page. Mishra says directing computing resources toward difficult document sections helps control processing costs.

Unsiloed currently employs seven people. The new capital will support additional research and engineering hires, expanded sales operations and development of proprietary training datasets.

The company’s next challenge is showing that its reported accuracy holds up across diverse enterprise documents. Reliable extraction, verifiable results and deployment costs will remain important considerations for customers adopting AI agents.

Shivangi Yadav

Shivangi Yadav is a technology writer at TechAmerica.ai, covering artificial intelligence, startups, digital platforms, consumer technology, mobility, and emerging technologies. Her reporting follows major developments across the global technology industry… All stories by Shivangi

Related articles

More on Unsiloed AI, Felicis, artificial intelligence.

See More - All AI stories

Newsletters

The day's tech news, in five minutes

Written by the reporters who cover each beat. No sponsored stories, and you can unsubscribe in one click.

Send me

By subscribing you agree to our Terms and Privacy policy.

Thank you

Your newsletter subscription has been received. Thank you for subscribing to TechAmerica.ai.