Projects

A production-ready SaaS platform that detects cue-sports objects (pool, snooker, billiards) in images using the YOLOv11 neural network, and generates natural-language explanations of what’s detected

A macro-driven portfolio optimisation framework built for the Strand Global Macro competition at King’s College London — a supervised, uncertainty-aware return model that rotates an equity portfolio across defensive, growth, and cyclical sectors through the business cycle.

A machine learning framework comparing ensemble regression methods on tabular data, evaluating linear models, tree-based ensembles, and gradient boosting approaches to maximise predictive performance.

An active learning pipeline demonstrating how the TypiClust (TPCRP) algorithm selects informative training samples through self-supervised representation learning, K-means clustering, and typicality-based selection, tested on CIFAR-10.