Ventures

Unsiloed AI raises $7.5 million seed round led by Felicis

The San Francisco startup, founded by two IIT Kharagpur classmates, turns messy documents into data that AI systems can read.

The 30-second read

  • Unsiloed AI raised a $7.5 million seed round led by Felicis.
  • Its parser scored 88.0 on olmOCR-Bench, but the company ran the test itself.
  • Voice-AI startup Bolna AI is a named customer.
Unsiloed AI raises $7.5 million seed round led by FelicisIllustration

Unsiloed AI has raised a $7.5 million seed round led by Felicis, a Silicon Valley venture capital firm. Felicis as lead investor was first reported by TechCrunch. Y Combinator, Orange Collective and Uncorrelated Ventures also joined. So did angels Aarthi Ramamurthy and Beth Turner. The company was part of the Fall 2025 Y Combinator batch, first reported by TechCrunch.

Unsiloed's software takes in PDFs, pictures and spreadsheets and outputs JSON and Markdown that LLMs and AI agents can use. It uses its own dual-stream vision models. Aman Mishra, co-founder and CEO, said: "The problem with LLMs is they’re inherently probabilistic". Adnan Abbas is the other co-founder, listed as CTO on the YC profile. The team has seven people. Pricing starts at one cent per page.

On olmOCR-Bench, a public benchmark by Allen AI, Unsiloed Parser scored 88.0. Reducto Agentic scored 66.0 and Unstructured 39.9. Unsiloed ran the tests itself, so the results are company-reported. The company also cites 99.78% accuracy in a customer test across roughly 84,000 data points in 542 financial documents. That figure is not independently verified.

The startup enters a crowded market with Reducto, Unstructured, Microsoft and Amazon. It plans to hire researchers and engineers, expand sales and build proprietary datasets, first reported by TechCrunch. Customers span finance, legal and healthcare, including Bolna AI.

Why it matters

Two IIT Kharagpur classmates are building core infrastructure that lets AI agents read business documents. Voice-AI startup Bolna AI already uses the product.

The data

olmOCR-Bench scores (company-run tests)
SystemScore
Unsiloed Parser88.0
Reducto Agentic66.0
Unstructured39.9

More in Ventures

More Ventures