I build systems where models have to work beyond the notebook.
I work on problems where model quality is only one part of the system. I care about how data is structured, how results are evaluated, and how the surrounding software behaves under real world constraints.
My current work ranges from leakage aware time series modeling in DeepFlood to extraction, retrieval, and asynchronous data processing in AppLynk. In both, I focus on making technical decisions measurable, inspectable, and defensible.
- Trustworthy machine learning and model evaluation
- Information retrieval and data systems
- Reliable software around probabilistic components
| Project | What it demonstrates |
|---|---|
| AppLynk | An AI-assisted opportunity intelligence platform combining LLM extraction, semantic search, ranking, and deduplication to help users discover relevant opportunities. |
| DeepFlood | A reproducible streamflow estimation system built around local feature engineering, time-series modeling, and leakage-aware evaluation. |
| Portfolio | A deeper look at my work, technical decisions, experience, and recognition. |
Programming: Python, TypeScript, SQL, C++, C
AI & data: TensorFlow, Keras, scikit-learn, pandas, NumPy, embeddings, vector search
Systems: PostgreSQL, pgvector, FastAPI, Next.js, Git, Linux
My background includes competitive programming, mathematics, and research communication. More context on experience and recognition is available on my portfolio.
I am open to conversations about AI/ML, data engineering, and software engineering opportunities, especially work where correctness, evaluation, and practical impact matter.
Based in Kowloon, Hong Kong · Vietnamese (native) · English (IELTS 7.5)