San Francisco-based Unsiloed AI has raised $7.5 million in seed funding to develop document-processing technology that makes information in business documents more accessible to AI agents. Silicon Valley venture capital firm Felicis led the round, with participation from Y Combinator, Orange Collective and Uncorrelated Ventures, as well as angel investors Aarthi Ramamurthy and Beth Turner.
The company is targeting the difficulty AI systems have interpreting tables, charts and complex layouts in business documents. Incorrectly extracted information can be costly for organizations such as banks, hospitals and law firms that rely on those documents in their daily work.
Unsiloed’s software processes documents, extracts information from tables, charts and other elements, and converts it into structured data that AI agents can use. It also shows where each piece of information came from in the original document, allowing customers to check the results.
Co-founder and CEO Aman Mishra said the inherently probabilistic nature of large language models can produce deviations and hallucinations with serious downstream consequences. Mishra founded the company with Adnan Abbas, his former classmate at the Indian Institute of Technology (IIT) Kharagpur. After spending time developing the technology, the pair joined Y Combinator’s fall 2025 batch.
Customers and competition
Mishra told TechCrunch that the startup already has customers in finance, legal services and healthcare. An unnamed bank uses the technology to process documents, while a law firm runs it on its own servers to handle contracts and old case files. Mishra also named voice AI startup Bolna AI as a customer, saying it uses Unsiloed to ingest knowledge bases for building voice agents.
Unsiloed has entered a crowded market. Competitors include specialist document-processing companies such as Reducto and Unstructured, as well as offerings from Microsoft and Amazon. Businesses also use general-purpose AI models to extract information from documents.
Processing text and images separately
Mishra said the company’s advantage lies in its approach to complex documents. Unsiloed uses what it calls dual-stream vision models, which handle text and image data separately. Another technique, called heat map-based chunking, helps identify the most important parts of a document.
The system then directs more computing power to areas such as figures, numbers and charts. A specialized decoder also takes industry terminology into account to distinguish between terms that have different meanings in different business contexts.
To support the company’s claims, Mishra pointed to a test involving 542 financial documents supplied by an enterprise customer. According to a technical brief shared by the startup, Unsiloed achieved 99.78% accuracy across roughly 84,000 extracted data points. The test compared it with major AI models, Microsoft Azure Document Intelligence and the customer’s own extraction tool, but the results have not been independently verified.
The startup also claims it outperformed competitors on olmOCR-Bench, a public benchmark developed by Allen AI. In its own tests, Unsiloed scored 88.0, compared with 66.0 for Reducto’s agentic system and 39.9 for Unstructured. The benchmark covered 1,403 PDF pages and examined how systems handled tables, mathematical expressions, scanned documents and complex layouts.
Mishra said Unsiloed’s pricing starts at one cent per page. He argued that the company also keeps costs down by directing computing power toward sections that need more attention rather than processing every part of a document in the same way.
The company currently has seven employees. It plans to use the new funding to hire more researchers and engineers, expand sales and develop proprietary datasets to improve its technology.



