Senior Data Scientist in Singapore. I build machine learning on large-scale time-series data, and I like taking models all the way from a messy dataset to something that holds up under external validation.
- 🔭 Currently at Respiree, building predictive monitoring models on continuous sensor streams
- 📈 Shipped a LightGBM model that beat the incumbent commercial benchmark by 46% on positive predictive value, using only 4 input signals
- 🗄️ Built pipelines over a 3.8M-subject dataset (16.2B time-series records) with SQL and Polars
- 🧠 Trained transformers, CNNs and ResNets from scratch in PyTorch and TensorFlow
- 🛠️ Outside work, I build small apps end to end: frontend, backend, deployment
| Project | What it is | Stack |
|---|---|---|
| Concertly | Personal concert diary with stats on every show I've been to (68 concerts since 2010). Code is private, the app is live | TanStack Start, Supabase |
| Nutrition Plan · live app | Bilingual (EN/FR) meal planner that hits daily macro targets. Parses free-text ingredients with an LLM and pulls nutrition values from a food database | Python, Streamlit, LLM API |
- Mayo Clinic Proceedings: Innovations, Quality & Outcomes (2025): external validation of a predictive monitoring model built on continuous sensor data. Read the paper
I write Alive and Well, a weekly newsletter on fitness, nutrition and evidence-based health. Read it on Substack
Most of my professional work lives in private repositories. Happy to talk through it in more detail.