Unsiloed AI Raises $7.5 Million in Seed Funding
The OCR startup, which converts documents into structured data, was led by Felicis and is a Y Combinator Fall 2025 company.
Updated on Nov. 20, 2026 in Startups

Unsiloed AI, an OCR startup that converts PDFs, scanned forms, presentations, Word documents and images into structured Markdown or JSON, has raised $7.5 million in seed funding. The company's parser scored 88.0 on olmOCR-Bench, ranking first among 19 systems tested.
Why it matters
The round reflects continued investor interest in tools that turn unstructured documents into data that applications can use and validate. Unsiloed's emphasis on confidence scores, source citations, and private-cloud or air-gapped deployment speaks to organizations that need auditable extraction for sensitive records.
The $7.5 million seed round is paired with an olmOCR-Bench score of 88.0, ranking first among 19 systems on a benchmark of 1,400 PDF pages and 7,000 binary unit tests. The extraction does not disclose valuation, revenue or customer figures.
The players
Unsiloed AI
An OCR startup that converts PDFs, scanned forms and images into structured Markdown or JSON, offering private-cloud, on-premises and air-gapped deployment options.
Felicis
A venture capital firm that led the seed round for Unsiloed AI.
Y Combinator
A startup accelerator that Unsiloed AI joined in its Fall 2025 batch and that also participated in the seed round.
The details
Unsiloed AI's API lets applications extract fields against custom schemas, classify and split documents, and edit text or images while preserving formatting. The platform uses a dual-stream vision architecture that processes content and visual structure together, combining deterministic computer vision algorithms with dual-stream vision models. The seed round was led by Felicis, with participation from Y Combinator, Orange Collective, Uncorrelated Ventures, Aarthi Ramamurthy and Beth Turner.
Timeline
2024-2025: Unsiloed AI is founded by Aman Mishra and Adnan Abbas.
Fall 2025: The company joins Y Combinator's batch.
November 20, 2026: The seed funding announcement is published.
Market Landscape
Unsiloed's 88.0 score on olmOCR-Bench, scored with Allen AI's unmodified deterministic scorer, places it first among 19 systems on that benchmark. The result positions the startup among a growing set of document-parsing tools competing on measurable extraction accuracy rather than general claims.
For organizations processing invoices, forms or contracts, tools like Unsiloed offer schema-based field extraction with confidence scores and source citations for checking results. Teams with data-residency requirements may find the private-cloud and air-gapped deployment options relevant to evaluating the product.
The takeaway
Document-parsing accuracy is increasingly measured with public benchmarks, so buyers evaluating extraction tools can compare vendors on tested results rather than marketing claims. Checking confidence scores and source citations against your own documents remains the practical next step.
Further reading
For more on early-stage company funding, see the Startups section.
Source note: This article includes information reported by Blocksandfiles.






