Software Engineering undergraduate at Wuhan University of Technology.
Building small, traceable proof-of-work projects in medical AI, AI agent evaluation, financial ML diagnostics, and reproducible ML experimentation.
- AI Agents and LLM Tool Use: tool-use evaluation, agent benchmarking, latency/cost analysis, and guardrail-aware diagnostics
- Medical AI: interpretable clinical data modeling, biomedical image analysis, robustness diagnostics, and clinically motivated ML evaluation
- ML for Decision-Making: representation learning, financial ML diagnostics, chronological evaluation, and leakage-aware model assessment
- Reinforcement Learning: sequential decision-making, baseline reproduction, and experiment-driven research training
- Reproducible ML Systems: environment repair, data-contract checks, checkpoint evaluation, runtime diagnostics, and technical documentation
Additional background: robotics simulation, GPU acceleration, radar / sensor signal processing, and optimization modeling.
| Direction | Proof of Work | Evidence |
|---|---|---|
| AI Agents | mini-agent-tooluse-benchmark | Controlled tool-use tasks, success / latency / cost metrics, invalid-call analysis, simple guardrail checks |
| Medical AI | NIPT Timing Optimization and Aneuploidy Screening | Interpretable clinical data modeling, optimization, imbalanced classification, SHAP-based analysis |
| Biomedical Image Analysis | cellseg-robustness-diagnostic | Segmentation robustness diagnostics, baseline evaluation, failure-case analysis, reproducible benchmark documentation |
| Financial ML | lob-representation-diagnostic | Representation diagnostics, chronological-vs-random split comparison, leakage-aware evaluation |
| Reproducible ML | HGSFusion | Environment repair, data-contract checks, checkpoint evaluation, training dry-runs, runtime diagnostics |
- GPU-Accelerated CFAR and MVDR Operators: OpenCL-based maritime radar signal-processing operators with memory-layout adaptation, workload tiling, and latency optimization.
- ROS 2 / Gazebo Autonomous Lunar Rover System: simulation-based autonomous navigation, vision recognition, mmWave radar perception, obstacle handling, and task-level validation.
ML and Data
Python · NumPy · pandas · scikit-learn · XGBoost · imbalanced-learn · SHAP · basic PyTorch
Modeling and Optimization
Regression · LASSO · GA · NSGA-II · Dynamic Programming · Monte Carlo Simulation · Sensitivity Analysis · PR/ROC Analysis
Systems and Robotics
Linux · Git · Docker · OpenCL · ROS 2 · Gazebo
Research Workflow
LaTeX · Markdown · reproducible experiment documentation · technical report writing · experiment logs · research memos