The South Korean document intelligence startup Upstage Co. Ltd. wants to expand into the U.S. after raising $45 million in a Series B bridge round.

Today’s round, led by the Korea Development Bank and the venture capital arms of Amazon.com Inc. and Advanced Micro Devices Inc., comes 16 months after Upstage announced the completion of a $72 million Series B investment. It brings the startup’s total amount raised so far to $157 million.

Upstage is the creator of a Document Intelligence suite that incorporates artificial intelligence features such as Document Parse, used to transform unstructured documents, including paper files, into structured formats that are ready for use by large language models. Another feature is Information Extract, which employs optical character recognition technology to identify and pull key data points from raw text.

The platform can process high-variation document packets such as policies, contracts, financial statements, invoices and emails with an impressive 99% accuracy. According to the company, that can substantially reduce the need for workers to spend time on manual data entry, while generating insights to help automate decision-making.

Upstage says that its platform is different to traditional OCR tools, which can only extract static information without any understanding of context. Instead, it looks at each document and interpret it in a holistic way, meaning that it can understand the layout, meaning and purpose of what it’s seeing. It’s uniquely able to recognize not just text, but also things like checkboxes and graphs, even on paper documents where the text is skewed, rotated or slanted, or where the paper itself is wrinkled.

Because of this, Upstage says its platform is ideal for any organization that has boxes stuffed full of old paper files waiting to be digitized. However, its primary focus is on the insurance industry, which it says is ripe for an AI-driven digital revolution.

It cites a recent study that found U.S. insurance companies spent $25.7 billion on adjudication costs last year, with $18 billion of that tied to overturned denials and unnecessary review cycles. The reason for this is that much of this work is performed manually by humans working with stacks of paper documents and fragmented systems, Upstage says. So this is where it wants to help.

“Most enterprise teams are still buried in document chaos,” said Upstage’s head of U.S. business Kasey Roh. “We bring structure and speed to that mess, helping carriers and brokers move faster, reduce costs and serve clients with confidence.”

The startup has already achieved a lot of success in South Korea, serving dozens of its top insurance firms as well as Samsung Electronics Co. Ltd. and a number of government agencies. Now, it has its eyes fixated on the U.S. market, and to seize that opportunity it has forged a close partnership with Amazon Web Services Inc., naming it as its preferred cloud provider.

Going forward, it’s hoping to improve its underlying Solar LLM family of AI models, which are the secret sauce that powers its Document Intelligence suite. The Solar models were built on the Amazon SageMaker platform and can be downloaded via the Amazon Bedrock marketplace. Upstage also plans to utilize that company’s machine learning infrastructure to enhance the quality of its models further and develop new agentic capabilities.

Upstage co-founder and Chief Executive Sung Kim said today’s round is a pivotal moment for the company and for enterprise generative AI.

“The next phase of AI is about performance, precision and readiness for real-world complexity, and this is exactly what we have built with our suite,” he said. “Our collaboration with AWS brings scale, credibility and deep technical alignment, which is critical as we deliver production-ready AI in high-stakes sectors like insurance.”

Images: Upstage