News General
15 views

IIT Kharagpur alumni’s Unsiloed AI raises $7.5 million to make business documents AI-ready

Nov 19, 2026 📍 Phliadelphia,PA, USA
IIT Kharagpur alumni’s Unsiloed AI raises $7.5 million to make business documents AI-ready
# Unsiloed AI Raises $7.5 Million to Help Businesses Make Documents AI-Ready

Unsiloed AI has secured $7.5 million in seed funding to develop technology that makes complex business documents easier for artificial intelligence systems and AI agents to understand. The funding round was led by Felicis, with participation from Y Combinator, Orange Collective and Uncorrelated Ventures, alongside angel investors Aarthi Ramamurthy and Beth Turner.

The startup is addressing a growing challenge for companies adopting AI tools: large language models often struggle to accurately interpret documents containing complicated tables, charts, figures and layouts. These limitations can lead to incorrect information being extracted and errors in the decisions or workflows that rely on it.

Unsiloed aims to solve this problem by converting complex documents into structured information that AI agents can process more reliably. Its software identifies and extracts content from different document elements, including tables and charts, while preserving links to the original sources. This allows users to trace extracted information back to the relevant section of a document and check whether the results are accurate.

Co-founder and CEO Aman Mishra said the probabilistic nature of large language models creates risks when businesses depend on their outputs for important operations. Even small variations or hallucinations can have significant consequences for downstream processes, particularly in industries where accuracy and verification are essential.

Mishra established Unsiloed with Adnan Abbas, a former classmate from the Indian Institute of Technology (IIT) Kharagpur. After developing the technology, the founders joined Y Combinator’s Fall 2025 batch, which helped support the startup’s early development.

The company is already working with customers across financial services, legal services and healthcare, according to Mishra. He said one bank uses the platform to process documents, while a law firm has installed the technology on its own servers to manage contracts and archived legal case files. He did not disclose the names of these two customers.

Mishra also identified Bolna AI, a voice artificial intelligence startup, as a customer. The company uses Unsiloed’s technology to process knowledge bases that support the development of voice-based AI agents.

## How Unsiloed Processes Complex Business Documents

Mishra said the startup’s technology is designed to distinguish itself through the way it analyses documents containing multiple types of information. Its system uses what the company describes as dual-stream vision models, which process text and image information separately before combining the results.

One stream handles text tokens, while the other examines image-based information from the document. This approach is intended to help the system understand both written content and visual elements rather than treating an entire document as a single block of text.

Unsiloed also uses a method called heat map-based chunking to identify the most important areas of a document. Instead of distributing computing resources equally across every page or section, the technology directs additional processing toward information-rich elements such as figures, numerical data and charts.

The startup has also developed a specialised decoder designed to recognise terminology used in particular industries. This is important because the same word or abbreviation can have different meanings depending on whether it appears in a financial, legal or healthcare document.

Together, these techniques are intended to improve the accuracy of information extraction while reducing unnecessary computing costs. The company believes this approach can make complex enterprise documents more useful in AI-powered business operations.

## Startup Reports 99.78% Accuracy in Financial Document Test

Mishra highlighted a test involving 542 financial documents provided by an enterprise customer. According to a technical brief shared by Unsiloed, the system achieved 99.78% accuracy across approximately 84,000 extracted data points.

The evaluation compared Unsiloed’s performance with several major AI models, Microsoft’s Azure Document Intelligence and the customer’s internal data-extraction system. However, the reported results have not been independently verified, and the figures were provided by the startup.

The company is positioning its technology as a solution for organisations that need AI systems to work with large volumes of business information without losing the context and structure contained in original documents.

Mishra said Unsiloed’s pricing begins at one cent per page. He explained that the company’s selective processing approach helps control expenses by allocating more computing power to sections that require detailed analysis instead of applying the same level of processing to every part of a document.

With its latest seed funding, Unsiloed is seeking to address a practical obstacle to wider enterprise AI adoption: making the information stored in business documents accessible, structured and verifiable for AI agents.
0 Upvotes
0 Downvotes
0 Likes

Login or register to upvote, downvote, and like this post.

Tags

news

Comments (0)

Login to post comments

No comments yet

Be the first to share your thoughts about this post.

Contact Information

Name: Nileena Sunil

Share This Post