東京工科大学大学院 コンピュータサイエンス専攻 修士2年（2027年3月修了見込み）｜NVIDIA Student Ambassador｜LLM × ロボティクス／開発者向けツール
M.Sc. candidate in Computer Science at Tokyo University of Technology. I build developer tools that people use, and on-device LLM systems for robots.
- NVIDIA AI Technology Center (NVAITC) Student Ambassador: four workshop talks
- IPSJ 第88回全国大会 口頭発表 (2026/3, first author)
- Kaggle Bronze Medal (team): March Machine Learning Mania 2025 (150/1727, top 8.7%)
- Languages: Chinese (native) · Japanese (educated in Japan from junior high through graduate school) · English (EIKEN Grade 2; reads papers and technical documentation)
| Area | Focus |
|---|---|
| Developer tools | Small, focused utilities in Go and Swift, shipped with releases and maintained against user issues |
| Systems | LLM x Robotics on Unitree Go2: voice, ASR, local LLM, safety compiler, robot control, all on a Jetson Orin NX |
| Research | Whether added model structure helps when data is limited: HRM-family two-state recurrent models against baselines matched in parameters and compute, on synthetic tasks (IPSJ 2026) and financial time series (master's thesis, ongoing; code not yet public) |
| Project | Description | Tech |
|---|---|---|
cc-clip  |
Paste images into remote Claude Code and Codex CLI over SSH; clipboard bridging for macOS and Windows | Go |
cmd-ime  |
Deterministic macOS input-source switcher: tap a modifier or press a shortcut to jump straight to any installed input method | Swift |
| Claudia | On-device natural-language control for the Unitree Go2: ASR, local LLM and a SafetyCompiler running on a Jetson Orin NX | Python / ROS2 Foxy / CycloneDDS |
| dotclaude | Collection of Claude Code skills for easy setup across different environments | Python |
| March ML Mania 2025 | Feature engineering, XGBoost model and evaluation for NCAA predictions (Kaggle bronze, team entry) | LightGBM / XGBoost / CatBoost |
