
たまご.exe
A portrait roguelike that combines slot drafting, score builds, and virtual-pet growth, culminating in a dragon evolution that breaks through the game UI.
- Godot 4.7
- GDScript
- PWA
- HTML5
- Automated Testing
Nihon University · Mathematical Information Engineering · Web & Game Development
Math × creativity — a developer who builds everything from the system up to the experience.

A portrait roguelike that combines slot drafting, score builds, and virtual-pet growth, culminating in a dragon evolution that breaks through the game UI.

A graduation research platform that compares LLM, reinforcement-learning, and human decisions inside a JRPG battle with lethal attacks and strict action-point constraints.

A WebGL interaction that tracks both hands and maps gestures to particle explosion, reconstruction, orbital rotation, and model switching.

A browser game combining kanji fusion, tower defense, and an endless roguelike. Joining characters such as 趙 and 雲 summons Zhao Yun, making written language itself the battlefield.

A geometric tower-defense roguelike where orbiting turrets and a controllable drone protect a central planet. Placement angles and rotating firing arcs continuously reshape the strategy.
INTERACTION LAB / 01
A place to test the site's reactions in small, light, and meaningful forms.
The ink quietly responds to the pointer.
Effects adapt to the motion preference of the device.

Started building web and game projects alongside coursework.
Supports lab members on the technical side of their projects.
Expected to graduate in 2027.
My strength is taking on unfamiliar problems, learning what I need on my own, and iterating until something real works. When I hit an unknown technology, I break the problem down, research it, get a minimal version running, and then improve it. In my seminar I also help members untangle problems in their own code.
Evaluating how AI agents and humans differ in decision-making under extreme conditions
Built a 3-vs-1 RPG battle environment and compared the decisions of humans, LLMs, and reinforcement learning, quantifying their distinct behaviors: RL optimizes survival, LLMs hallucinate rules, and humans take intuitive risks.
Computational Electromagnetics Lab, Nihon University
For work or internship inquiries, feel free to reach out.