Patronus AI has raised $50 million in Series B funding, led by Greenfield Partners, with participation from Lightspeed Venture Partners, Notable Capital, Datadog, and Samsung Ventures. The round brings the startup’s total funding to $70 million, according to the company. Founded by former Meta AI researchers Anand Kannappan and Rebecca Qian, Patronus aims to build infrastructure that lets developers train AI agents in simulated environments, ensuring they can perform reliably in real-world scenarios.
The startup focuses on creating digital replicas of websites and corporate applications, where AI agents can be stress-tested using reinforcement learning. This technique rewards agents for completing tasks and penalizes them for failure, allowing them to adapt to unpredictable situations. Kannappan said benchmarks alone aren’t enough to gauge an agent’s ability to handle ambiguity or recover from errors. “They don’t tell you whether an agent can handle ambiguity or operate reliably across long, unpredictable workflows,” he said.
Related: Warp lands $60M to automate payroll and HR
Patronus’s simulations are used by major AI labs and startups, with demand for its tools growing sharply. The simulations mimic real-world conditions, letting developers test agents in scenarios ranging from financial trading to healthcare diagnostics.
The company currently targets finance and software engineering tasks but plans to expand into areas where verification is harder. Kannappan said the market is uncrowded, with few rivals matching Patronus’s agentic testing capabilities. Competitors like Google and Decart AI focus more on training than evaluation, leaving a gap Patronus aims to fill.
infrastructure problems in AI are central to the company’s mission. As autonomous systems grow more complex, the need for reliable simulations becomes critical. Patronus’s tools let developers practice in environments that mirror real-world unpredictability, much like how Waymo uses simulations to train self-driving cars.
Related: Ford’s Electric Car Pioneer Exits Amid Shift in EV Strategy
For now, the startup remains focused on verifiable problems, such as financial transactions or software workflows. But Kannappan hinted at broader ambitions, acknowledging many areas remain difficult to test. “There are a ton more areas that are very non-verifiable,” he said, pointing to challenges in fields like social interaction or creative problem-solving.
The funding will accelerate Patronus’s growth as AI systems take on more complex tasks. From booking restaurant tables to managing stock trades, autonomy is expanding—but so are the risks. Patronus’s simulations aim to bridge the gap between lab benchmarks and real-world performance, ensuring agents can handle the unexpected without failure.
