Applied AI/ML and backend engineering work, ranked by relevance.
A pre-registered experiment measuring whether decomposing a compliance agent into a multi-agent system actually improves decision repeatability, across 41,400 scored runs and five open-weight model families.
Trained and compared XGBoost, Random Forest, and Logistic Regression on a 250,000+ record public health dataset, tuning decision thresholds to maximize recall for high-stakes screening.
Built and evaluated a CNN for static ASL gesture recognition against a Random Forest baseline using the Sign Language MNIST dataset, achieving 99.26% accuracy.
A modular, layered tool for ingesting, cleaning, querying, and visualising health-related CSV data over SQLite, built with test-driven development throughout.
Backend for a multi-merchant e-commerce marketplace, built and led as co-founder and lead backend engineer.
Administration tools managing student records, course enrollment, and faculty operations for 5,000+ students across five faculties.