Wednesday, November 18, 2026
Noti Group
Technology

Artificial Intelligence Firm Unsiloed Raises $7.5 Million in Seed Funding

The funding will be used by Unsiloed AI to further develop its software that breaks down complex documents for AI systems.

Felicis leads $7.5M bet on Unsiloed AI to make enterprise documents usable by AI agents
Source: TechCrunch

Businesses are increasingly turning to artificial intelligence (AI) agents to automate work processes, but a significant challenge remains: making complex documents accessible to these agents. This is where Unsiloed AI comes in, having raised $7.5 million in seed funding to tackle the problem.

The funding round was led by Silicon Valley venture capital firm Felicis, with participation from Y Combinator, Orange Collective, and Uncorrelated Ventures. Additionally, angel investors Aarthi Ramamurthy and Beth Turner also invested in the company. The funding will be used to further develop Unsiloed's technology.

Complex documents often contain tables, charts, and layouts that AI systems struggle to interpret accurately. This can lead to costly mistakes for businesses, with severe consequences in industries such as banking, healthcare, or law, which rely heavily on these documents for their daily operations.

Unsiloed aims to solve this problem by developing software that breaks down complex documents, extracts information from tables and charts, and turns it into structured data that AI agents can use. The technology also provides a clear audit trail of the extracted information, allowing businesses to verify the results rather than relying solely on the AI's output.

Unsiloed's technology aims to address a significant issue in Large Language Models (LLMs), which are inherently probabilistic and prone to errors such as hallucinations. This can have far-reaching consequences downstream, according to Aman Mishra, the company's co-founder and CEO.

Mishra and his colleague Adnan Abbas developed Unsiloed after meeting at the Indian Institute of Technology (IIT) Kharagpur. They spent time refining their technology before joining Y Combinator's Fall 2025 batch. This experience has helped them fine-tune their approach to document processing, which is a critical aspect of AI adoption in various industries.

Unsiloed is already working with customers across finance, legal services, and healthcare. For instance, one bank is using the technology to process documents more efficiently. A law firm has also deployed Unsiloed on its servers to handle contracts and old case files.

Unsiloed's innovative approach to processing complex documents involves a unique "dual-stream vision model that handles text and image data separately. This technique allows for more accurate identification of key information within documents, which is particularly useful in enterprise settings where multiple stakeholders may need access to the same data.

The startup's use of heat map-based chunking" is another crucial aspect of its technology. By analyzing the most important parts of a document and directing computing power toward them, Unsiloed can quickly identify figures, numbers, and charts that require attention. This capability enables AI agents to extract relevant information from documents more efficiently.

Industry-specific terminology poses a significant challenge for many AI models, but Unsiloed has developed a specialized decoder to address this issue. The decoder takes into account the nuances of industry-specific language, allowing it to distinguish between terms with multiple meanings in different business contexts. This level of precision is essential for accurate data extraction and analysis.

A recent test involving 542 financial documents supplied by an enterprise customer demonstrated Unsiloed's capabilities. According to a technical brief shared by the startup, the technology achieved 99.78% accuracy across approximately 84,000 extracted data points. The results were compared to those of major AI models, including Microsoft's Azure Document Intelligence and the customer's in-house extraction tool.

Unsiloed's AI document-processing tool has demonstrated impressive performance on olmOCR-Bench, a public benchmark developed by Allen AI. In trials conducted by the startup, Unsiloed scored 88.0, outpacing other AI systems in the test. These results were compared to those of major AI models, including Microsoft's Azure Document Intelligence and an internal extraction tool used by the customer.

The benchmark covered a substantial amount of data - 1,403 PDF pages - and evaluated how well the systems handled various document types. Unsiloed's performance was notable in its ability to process tables, mathematical expressions, scanned documents, and complex layouts with ease.

Unsiloed's pricing strategy is designed to be cost-effective for businesses, starting at a low rate of one cent per page. By directing computing power towards the most critical areas of a document, Unsiloed claims to reduce costs and improve efficiency in its processing methods.

The startup currently has a small team of seven people, but plans to use the fresh funding to expand its workforce. New hires will include researchers and engineers who will help develop proprietary datasets to enhance the technology's capabilities.

Felicis Ventures has led a $7.5 million investment in Unsiloed AI, a company working on enabling enterprise documents to be usable by artificial intelligence agents.

This funding round will allow Unsiloed AI to scale its operations and develop more advanced technologies that can extract insights from complex data sets stored in enterprise documents. The company's proprietary datasets are expected to enhance the capabilities of its AI technology, making it even more effective at extracting valuable information from large volumes of unstructured data.

Unsiloed AI aims to make a significant impact on industries such as healthcare, finance, and law by providing a solution for unlocking insights from existing documents, rather than requiring manual processing or reformatting.

Facts based on reporting originally published by TechCrunch.

You may republish this story, in full or in part, if you credit Noti Group and link to it (licence CC BY 4.0). Photos are not included.