Proximal
Proximal builds RL environments for software engineering. Our environments are grounded in real-world codebases and simulate the range of complexity within real software engineering tasks. We sell these environments to frontier labs and AI startups to help them improve their models and coding agents.
We’ve only scratched the surface of what’s possible in automating software engineering. Coding agents are powerful, but far from being good at long-horizon autonomous coding. We close this gap by giving labs high-quality coding environments to train on.
Team
Justus Mattern: Led RL research & data at Prime Intellect. He was a core contributor to their RL training framework prime-rl, their reasoning datasets Synthetic-1 & Synthetic-2 and their latest reasoning model Intellect-2.
Navid Pour: Was the second founding engineer at Cursor. He was part of the original team that took Cursor tab from an idea into production and co-led multiple experiments on coding agents. Also built infrastructure for browser agents at Browserbase.
Calvin Chen: Built a logistics software company that he bootstrapped to $1.5M ARR and sold for $9M. Most recently he cofounded Fetchr with Navid, building AI agents for shopping.
For inquiries, you can reach us at contact@proximal.so.