Unsiloed AI Secures $7.5 Million to Bridge Data Gap for AI Agents
San Francisco startup Unsiloed AI raised $7.5 million in seed funding to help enterprise AI agents accurately extract data from complex documents like charts and tables.
🤖AI ModelsPhoto: TechCrunch AI
Key points
- Unsiloed AI raised $7.5 million in a seed round led by Felicis to improve how AI agents interpret complex documents.
- The technology uses a unique 'dual-stream' vision model to process text and images independently, aiming to reduce hallucinations.
- The startup claims near-perfect accuracy in financial data extraction and is targeting the legal, healthcare, and banking sectors.
San Francisco-based startup Unsiloed AI has raised $7.5 million in seed funding to solve the "unstructured data" problem that often prevents AI agents from accurately reading complex business documents. Led by venture firm Felicis, with participation from Y Combinator and Uncorrelated Ventures, the investment aims to scale technology that converts intricate charts, tables, and layouts into machine-readable data for high-stakes industries.
The startup, founded by Aman Mishra and Adnan Abbas, addresses the inherent limitations of Large Language Models (LLMs) which can struggle with visual layouts. Because LLMs are probabilistic and prone to hallucinations when interpreting visual data, Unsiloed uses a "dual-stream" architecture. This system processes text and image data separately to ensure higher precision and provides citations so users can verify the source of extracted information.
Unlike general-purpose models, Unsiloed’s software employs a technique called "heat map-based chunking" to prioritize computing power on critical document elements like figures and mathematical expressions. This specialized approach is designed to handle industry-specific terminology across the finance, legal, and healthcare sectors, where errors in document processing can lead to significant operational risks.
The company reports early traction with clients including a bank, a law firm, and the voice AI firm Bolna AI. In internal testing involving over 500 financial documents, the startup claimed a 99.78% accuracy rate across 84,000 data points. However, these performance figures have not been independently verified. Unsiloed enters a competitive landscape that includes tech giants like Microsoft and Amazon, as well as specialized startups like Reducto and Unstructured.
What to watch next: Look for whether Unsiloed can maintain its reported accuracy edge as it scales its "one cent per page" pricing model across high-volume enterprise workflows.
Source TechCrunch AI — techcrunch.com/2026/11/18/felicis-leads-7-5m-bet-on-unsiloed-ai-to-make… ↗
Read next
🧠AI AnalysisSalesforce Optimizes Global AI Model Inference with Amazon SageMaker
Salesforce has implemented cross-Region inference routing in Amazon SageMaker to manage AI model capacity across 30+ global regions more efficiently.
🏢Big TechMicrosoft Challenges Rivals on AI Pricing at Ignite Conference
Microsoft is positioning its Copilot AI as the more cost-effective choice for businesses compared to OpenAI and Anthropic, citing a commissioned study to support its clai…
🚀Startups & FundingOpenAI CEO Sam Altman Addresses Upcoming Biographical Film
OpenAI CEO Sam Altman commented on the upcoming film 'Artificial,' which portrays his life and the 2023 leadership crisis at the company.
🔌Chips & HardwareSnapchat Integrates AI Coding Assistants into AR Lens Development
Snapchat has launched the Easy Lens CLI, enabling developers to build augmented reality experiences using AI coding tools like Claude Code and Cursor.
🚀Startups & FundingRadixArk Unveils DaisyRose-27B, Focused on Agentic Efficiency
Startup RadixArk has launched DaisyRose-27B, an AI model optimized for long-horizon agent tasks, claiming significant improvements in training efficiency and reasoning ca…
🔌Chips & HardwareUS Eyes the Philippines for Massive AI and Chip Hub Amid Domestic Opposition
President Donald Trump visited the Philippines to advance the Pax Silica initiative, proposing a 1,620-hectare hub for data centers and chip fabs to bypass local pushback…