Discernis Raises $2.5M for AI That Finds Key Evidence Buried Across Millions of Documents

NEW YORK, NY, UNITED STATES, August 25, 2026 /EINPresswire.com/ — Discernis, legal intelligence software for litigation and investigations, today announced it has raised a $2.5 million seed round. The round was led by Newfund Capital, with participation from Triple Impact Capital, Remarkable Ventures, and C2 Ventures.

“Very few people in the world have actually built the training pipelines behind modern AI, and almost none of them are applying it to a market as demanding as litigation. Rob has the technical depth to run frontier-level AI entirely inside a law firm’s walls, and the judgment to know exactly where it has to run,” said Henri Deshays, General Partner at Newfund Capital.

Instead of sampling a document collection, Discernis AI reads every document and provides legal teams with complete answers to their questions. Discernis is in use today in Am Law firms, where litigation teams rely on it for active matters. The funding will support continued development of Discernis AI and expansion across large firms and corporate legal departments.

“The way we think about our AI is as one extremely smart associate who has read all of the documents in the case,” said Rob Kwiatkowski, founder and CEO of Discernis. “We’ve never had a client come to us and say: ‘Take your time.’ It’s always: ‘We need these documents reviewed by tomorrow, and there are fifteen million of them. Can you do that?’ The short answer is yes.”

Most legal AI can’t see beyond just a few documents. Some tools read each document in isolation, blind to its connections with the rest of the record. Others work like search: they retrieve a handful of documents that look relevant and reason over only those, leaving the rest of the collection unread. Discernis AI is purpose-built for the opposite: full analysis of entire collections, often millions of documents, and finds the evidence that only emerges across documents. Discernis reviews more than 50,000 documents per hour and scales with additional hardware.

Because Discernis already knows every document in the collection, attorneys don’t need to learn prompt engineering. Teams paste in the review protocol they already wrote, and Discernis works from it.

For regulated industries, government investigations, and matters governed by outside counsel guidelines, keeping client data off third-party AI services has become a procurement requirement, not a preference. Discernis was built for exactly this with a sovereign AI approach: the system deploys on-premises or in the customer’s own secure cloud environment, and all inference runs within it, with no third-party or commercial AI API called at any point.

Kwiatkowski holds a PhD in computer science from Columbia University, where his research on self-aware robotic systems appeared on the cover of Science. At AWS, he built Amazon’s first RLHF training pipeline, the technique behind modern AI assistants. He founded Discernis to apply the same technique to create purpose-built AI that finds the evidence everyone else misses. Discernis is based in New York City.

About Discernis
Discernis is legal intelligence software for litigation and investigations. It reads entire document collections using cross-document analysis to surface evidence and inform case strategy.

Rob Kwiatkowski
Discernis
rob@discernis.ai

Media gallery