Param Kumar

Param Kumar is co-founder and CEO of AgileRL, the London-based company making reinforcement learning accessible to any enterprise. A former product manager at Just Eat and PolyAI, Kumar launched AgileRL through Entrepreneur First in 2023 and has since grown its open-source framework to 300,000+ downloads.
This profile isn't ready yet! Check back soon.

Param Kumar is the co-founder and CEO of AgileRL, a London-based machine learning company he launched in 2023 alongside co-founder and CTO Nicholas Ustaran-Anderegg, with early backing from Entrepreneur First. Kumar's career in technology began at Just Eat Takeaway.com, where he worked as a Product Manager from 2019 to 2022 before moving to PolyAI as Senior Product Manager. It was during his time building AI products and, by his own account, attempting to construct a reinforcement learning system from scratch at a previous company, that he identified the core problem AgileRL was founded to solve: RL is one of the most powerful AI training techniques available, but the operational cost of deploying it in an enterprise context is prohibitive for all but the largest technology companies.

Kumar joined Entrepreneur First in September 2022 as a Founder-in-Residence, where he met Ustaran-Anderegg and developed the AgileRL concept. Since founding AgileRL in January 2023, he has led the company to release an open-source RL framework that has been downloaded over 300,000 times, and to build Arena — a managed RLOps platform used by customers including MIT, Carnegie Mellon, Roblox, IBM, Airbus, and JPMorgan. In January 2026, Kumar led a £6 million seed fundraise led by Fusion Fund, with participation from Flying Fish, Octopus Ventures, Entrepreneur First, and Counterview Capital. The company is now opening a San Francisco office to pursue enterprise RL demand in defence, robotics, and finance. Kumar's founding insight — that no company should need to build an AI research lab just to use reinforcement learning — has positioned AgileRL as one of the most distinctive infrastructure plays in the enterprise AI stack.

Is this you? Would you like to update some of the details here?

Features: