Patronus raises $50M Series B led by Greenfield Partners

What's the deal? Patronus, a San Francisco startup building simulated environments to test AI agents, has raised a $50 million Series B led by Greenfield PartnersDealroom has a profile for this one. Try Dealroom →.

Notable Capital, Lightspeed, Datadog, and Samsung also joined the round, which brings total funding to $70 million.

Founded in 2023 by former Meta AI researchers Anand Kannappan and Rebecca Qian, Patronus creates "digital world models" — replicas of websites and internal systems where agents are stress-tested using reinforcement learning.

Why now? AI agents are evolving from answering questions to autonomously executing complex, multi-step tasks like booking trips or running financial analysis.

But a high benchmark score doesn't prove an agent can do real-world jobs correctly. Model makers want proof before trusting agents with live tasks.

Demand is surging. Patronus' revenue has grown 15-fold over the past year, and virtually every frontier AI lab plus many startups are now customers, according to Notable Capital's Glenn Solomon.

What could go wrong? Agents tend to take shortcuts, failing to complete tasks correctly.

"Patronus is really good at spotting the hacks and making sure they are holding the models accountable," Solomon said.

The company currently covers software engineering and finance — problems it can immediately verify. Harder, non-verifiable areas remain a challenge.

Patronus also competes against the in-house evaluation teams AI labs have already built.

The signal: The participation of corporate backers like Datadog, Lightspeed, and Samsung alongside lead investor Greenfield Partners underscores how testing infrastructure is becoming strategically vital as agents move into production. For an early-stage company, landing nearly every frontier AI lab as a customer while growing revenue 15-fold suggests agent evaluation is emerging as a market in its own right.

Image credit: jurvetson