Training environments
for AI.

We build games and simulations that help models
learn to solve problems they haven’t seen before.

Research

Which experiences make a model more capable? We’re starting with interactive worlds, measuring what transfers to other tasks, and using those results to design better training environments.

Team

Anshul Gokul

CS & mathematics at Georgia Tech. Previously post-training research at DeepGrove.

Neil Mehra

CS at Georgia Tech. Previously quantitative trading systems at Optiver and Qube Research.