Document processing startup Extend said today it has raised $17 million via two separate early-stage funding rounds – a seed investment and a Series A round.

Innovation Endeavors was the lead investor, while the rounds also saw participation from Y Combinator, Character VC and angels such as the former Adobe Inc. Chief Security Officer Scott Belsky, Vercel Inc. Chief Executive Guillermo Rauch and Stripe Inc. product chief Jeff Weinstein.

Extend, officially known as CrowdView Inc., is trying to improve the speed and accuracy of document processing with artificial intelligence. The startup has developed specialized large language models that can transform complex PDFs and other files into validated, production-ready data. With its advanced LLMs trained to extract, classify and split information from documents, Extend says it can process documents with greater than 95% accuracy, no matter if it’s a clean PDF or a degraded scan or photocopy of a photocopy, understanding handwriting, tables, signatures and more.

The startup is trying to displace existing document processing tools that rely on a much older form of AI known as optical character recognition or OCR, which is known to struggle with handwritten notes and tables and cannot deal with images. It all starts with its ingestion pipelines, which leverages multimodal LLMs to break down the layout of any document to understand its content with high accuracy, semantically chunking elements together to create clean outputs that humans and machines can easily understand. Extend says its customers play a role in this, helping to fine-tune the capabilities of its LLMs to get better at understanding the intricacies of their documents.

As part of the document scanning process, Extend will also incorporate corrections to fix any errors it finds and improve the accuracy of each document, while its built-in evaluation tools give customers a way to analyze and evaluate its performance. Its software can be integrated directly into any business process or workflow, so employees in every department can analyze and generate insights from fresh data instantly.

Extend co-founder and CEO Kushal Byatnal said that the most critical data for a significant majority of businesses lives in their documents, which have historically always been challenging to access at scale. “We built Extend to make document processing as easy and as accurate as possible, giving technical teams modern infrastructure and non-technical users intuitive tools to improve accuracy in real-time,” he said.

Alongside today’s funding, the company is expanding its product lineup with a new, self-serve version of its document processing platform that can be easily deployed by nontechnical users, enabling anyone to set up a custom document automation workflow in days.

Byatnal said today’s funding will help the company to unlock the potential of unstructured data in almost any business context. It plans to use the funds to grow its engineering and go-to-market teams and invest more in research and development, with the aim being to accelerate its growth while simultaneously improving its product.

“We’ve seen remarkable momentum over the past year,” he added. “Now, we’re scaling to meet growing demand from technical teams who need accuracy, speed and flexibility in document automation.”

Image: Extend