AI startup Patronus AI raises $50 million to test agents in digital replicas

Patronus AI, a startup established by former Meta AI researchers Anand Kannappan and Rebecca Qian in 2023, has secured $50 million in Series B funding. This latest investment, led by Greenfield Partners with participation from Notable Capital, Lightspeed, Datadog, and Samsung, brings the company's total funding to $70 million.

The company specializes in creating "digital world models" – synthetic replicas of real-world systems like websites and internal business tools. These simulated environments allow AI agents to undergo reinforcement learning, where successful tasks are rewarded and failures are penalized. Patronus AI claims this approach, akin to how Waymo trained self-driving cars in virtual settings, can identify errors that traditional benchmarks overlook.

Patronus AI reports a 15-fold increase in revenue over the past year, attracting nearly every major AI lab and numerous emerging companies as clients. Currently, the startup offers simulated environments tailored for software engineering and finance sectors. Future plans include expanding capabilities to support AI agents performing longer-duration tasks, potentially operating for days or even weeks.

The company views its primary competition not as other "human-in-the-loop" firms, but rather as the internal evaluation teams already maintained by AI laboratories. This distinction arises because Patronus AI's simulations involve no direct human intervention during the testing phase.

“

The company says the approach, which draws comparisons to how Waymo trained self-driving cars in virtual environments, catches errors that conventional benchmarks miss.