Junming ShunmeiCho

About Me

Shunmei Cho (趙俊茗) — M.Sc. Candidate in Computer Science @ Tokyo University of Technology

"Explainable, Reproducible, Deployable."

Area	Focus
Research	Hierarchical Reasoning Models for small-data inference & structured knowledge under data constraints
Systems	LLM x Robotics (Unitree GO2) — Voice → ASR → LLM → Robot Control → Feedback Loop
Engineering	Reproducible ML experiments — Config-driven, W&B tracked, Dockerized

Small-Data Inference under Constraints

Methodology: Config-driven experiments | Clear baselines | Honest failure analysis | W&B tracking

Unitree GO2 End-to-End Prototype

Voice -> ASR -> LLM Reasoning
  -> Action Command -> ROS2/SDK
    -> Execution -> Feedback Loop

Key Concerns: Interface contracts | Latency budgets | Safety filters | Fallback behaviors

Project	Description	Tech	Highlight
Claudia	Intelligent quadruped robot with LLM for dialogue, voice and navigation	ROS2 / Unitree SDK / ASR	End-to-end LLM x Robotics
March ML Mania 2025	Feature engineering + ensemble pipeline for NCAA predictions	LightGBM / XGBoost / CatBoost	Top 8.7% (150/1727)
JetLLM	LLM deployment and inference optimization	CUDA / Python	GPU-accelerated inference
PEFT-SFT-Curator	Parameter-efficient fine-tuning and data curation pipeline	PyTorch / PEFT	Efficient LLM training