Data Science · Machine Learning · ML Systems

I turn messy data and model outputs into tested experiments, reliable pipelines, and useful products. I'm a Computer Engineering student at Nanyang Technological University (NTU) and currently a Data Scientist Intern at Vertex Holdings, working on applied data science and AI-enabled internal tools.

Previously, as a Machine Learning & GenAI Engineer Intern at Panasonic R&D Centre Singapore, I built local computer-vision and vision-language systems for GPU-constrained environments.

Python LightGBM scikit-learn pandas SciPy · Private team submission · early public prototype

• Co-built the original NAISC 2026 Singtel Challenge pipeline for telco churn, combining PSI/KS drift tests, adversarial validation, feature engineering, and reweighting under fixed-hyperparameter and CPU-only constraints.
• Built a self-contained drift dashboard and report generator; the final private team repository records 0.787 public-test AU-PRC.

PyTorch CLIP Whisper FAISS FastAPI Celery

• Implemented and trained fusion, two-tower retrieval, DeepFM ranking, and MMoE multi-task ranking components, then wired their checkpoints into an asynchronous serving path.
• Evaluated retrieval against content-similarity and popularity baselines, diagnosed where personalization underperformed, and fixed a hand-rolled AUC implementation using a known-answer regression test.

Python pandas SciPy SQLAlchemy FastAPI pytest

• Built an event-driven research backtester with next-bar execution, persistent orders, walk-forward analysis, and finite-sample-corrected permutation tests.
• Added quality-gated market-data ingestion, immutable data generations, and 11 SQL analytics/integrity queries; the public system passes 712 tests, Ruff, and strict mypy.

Qiskit scikit-learn Python Spring Boot

• Primary repository author in a two-person hackathon team, building a simulator-verified QSVM with quantum-kernel feature encoding over an imbalanced credit-card fraud dataset.
• Added precision/recall, F1, ROC-AUC, average-precision, and confusion-matrix evaluation, plus a typed Spring Boot service layer.

• FinancePy PR #249 — fixed Python 3.10/3.11 compatibility errors in bond YTM/OAS pricing code, including a second latent instance beyond the original report; merged upstream in August 2026.

Data & ML: Python SQL PyTorch scikit-learn LightGBM pandas NumPy SciPy FAISS

Evaluation: experiment design, baseline comparison, imbalanced classification, drift detection, walk-forward validation, permutation testing

ML Systems: FastAPI Docker Celery Redis Prometheus MLflow PostgreSQL

Engineering: Git pytest Ruff mypy C++ Java TypeScript

Nanyang Technological University, Singapore · Bachelor of Engineering (Honours), Computer Engineering · 2024–2028