Data Science & Analytics graduate from NUS. I like finding problems in data and building things to fix them — whether that's a product feature, a classifier, or a cleaner way to see the numbers.
Student Wellbeing Risk Platform
Built a Random Forest classifier to predict student mental health risk, with SHAP explainability and a React + FastAPI dashboard for teachers to act on predictions.
Python scikit-learn SHAP FastAPI React
Stress Detection in Reddit Posts
Compared three NLP approaches (LIWC features, TF-IDF, DistilBERT) for classifying stress in social media text. Fine-tuned DistilBERT achieved F1 = 0.82. Includes error analysis and a deployment risk assessment for real-world mental health triage.
Python DistilBERT scikit-learn LIWC
Epidemic Parameter Inference on Adaptive Networks
Used Rejection ABC and ABC-SMC to jointly infer infection and rewiring rates on an adaptive network epidemic model. Found that the β–ρ correlation is a structural non-identifiability — no sampler can resolve it without more informative data.
Python Approximate Bayesian Computation Sequential Monte Carlo
Gameplan
A Chrome extension I built to solve my own focus problem — task management and a Pomodoro-style timer in one minimal popup.
JavaScript Chrome Extensions API