Kohei Ozaki (smly)
Software Engineer
Google Developer Expert (Kaggle)
Game AI
RiichiEnv: Research Environment for Mahjong AI
A high-performance research environment for Riichi Mahjong AI development. The core game logic is implemented in Rust for fast state transitions and rollouts, with a Python Gym-style API designed for reinforcement learning. Features include compatibility with the Mortal bot via the MJAI protocol, flexible rule configuration (Tenhou/MJSoul presets, 3-player/4-player modes), and an integrated replay viewer for Jupyter Notebooks.
RiichiLab: Mahjong AI Competition Platform
An online competition platform where developers submit Mahjong AI bots that play full
four-player Riichi Mahjong games against each other. Participants write a Python program (bot.py) that communicates through the mjai protocol, a
standardized message-based interface for Mahjong AI. Each submission runs inside a
Docker-sandboxed environment, and a high-performance Rust game simulator handles state
management, rule enforcement, and scoring. The platform provides an Elo-based leaderboard,
detailed game logs, and a pip-installable local testing tool (pip install mjai)
for rapid iteration before submission.