Writing

Undergraduate thesis at TU Dublin: an explainable Graph Neural Network for anti-money-laundering on the Elliptic2 dataset, evaluated against the EU AI Act and the ALTAI seven-pillar framework. A two-layer GATv2 with attention pooling reaches 0.515 PR-AUC, roughly 2.47x the best published structure-only baseline.

Python / PyTorch Geometric / GATv2 / GNNExplainer / Integrated Gradients / Kernel SHAP

Designed and led the technical architecture for an AI-enhanced planning intelligence platform focused on the Irish planning system. Built a traceable, explainable AI workspace that combines planning law, policy, precedent and document workflows to reduce validation errors, research overhead and professional risk in regulated decision-making.

TypeScript / Golang / React / SQL / Document Pipelines / RAG