Second-year Data Science student at the University of Information Technology (UIT), VNU-HCM, based in Đà Nẵng, Vietnam.
I like problems where the data fights back: leakage, distribution shift, noisy labels, strict competition rules. I try to measure each idea with honest validation before I believe it.
- 🔭 Currently: preparing the finals of AISC 2026 (December 2026) with my team, and working on research on small language models for robot decision-making
- 🌱 Interests: tabular machine learning, Vietnamese NLP, small language models, robot learning
- 📊 Kaggle: kaggle.com/tdn2708
| Project | What it is | Highlights |
|---|---|---|
| Customer Radar | Multi-channel customer-feedback analytics with a Data Trust Layer that down-weights fake and paid reviews before any metric is computed | AISC 2026 finalist (team project) · fine-tuned ViSoBERT running in the browser via ONNX · live demo |
| Physical AI | Robot navigation in ROS 2 / Gazebo with a small language model behind a safety layer, plus a CNN + GRU lidar policy trained with DAgger | 16/16 Gazebo maze runs, 0 wall contacts in 4,384 test episodes |
| Airline Satisfaction | Kaggle Playground S6E10, ROC AUC | Public LB rank 134 / 3,735 (top 4%) · 278-model stack checked with nested CV |
Data science and ML: Python, pandas, NumPy, scikit-learn, LightGBM, XGBoost, CatBoost, PyTorch, Hugging Face Transformers, TabPFN / TabICL, Optuna
NLP and LLMs: ViSoBERT, LoRA fine-tuning, Ollama, ONNX Runtime
Robotics: ROS 2 Humble, Gazebo
Web and data: JavaScript, React, Node.js, MongoDB, SQL
Tools: Git, Linux / WSL2, Jupyter, Google Colab, Kaggle