Sotetsu Koyamada

Sotetsu Koyamada is a researcher and engineer interested in reinforcement learning (RL), multi-agent systems, and game AI. I have been involved in the following projects:

• Led the Japanese translation of the theoretical RL textbook by C. Szepesvári.
• Suphx: Microsoft’s Mahjong AI, which achieved an estimated rating higher than 99.99% of human players on an online Mahjong platform.
• Pgx: A scalable, ultra-fast suite of game simulators running on GPUs, written in JAX.
• Mjx: A high-speed Mahjong simulator written in C++/Python.
• Developed Bridge AI with state-of-the-art performance in the bidding benchmark against WBridge5.