I am an MComp Computer Science graduate from Middlesex University London (Upper Second-Class Honours). My interests are in trustworthy and responsible AI, machine-learning evaluation, uncertainty quantification, reproducible research, and computational sustainability.

I build transparent research prototypes that preserve uncertainty, document limitations, and make their evidence boundaries visible. I report what I find — including negative results.

Evidence-constrained credit-risk classification with calibration, drift/OOD detection, explainability, locked precedence rules, and mandatory human review. 76 automated tests, authenticated governance API, reproducible experiment ledger.

Physics-constrained least-cost abatement optimiser with Monte Carlo rank-reversal robustness analysis across electricity, heating, transport, waste and industry. Real-data validation against EPA eGRID natural-gas plants.

Electricity-demand forecasting for hotel venues with deliberate temporal and spatial generalisation tests. Locked future-period holdout; honest reporting of degradation under distribution shift.

A synthesis asking whether a single Robustness Retention Ratio can compare the three projects above. It cannot: the ranking reversed and the range widened more than four times when an equally defensible alternative comparison was substituted. Reported rather than buried.

Does FedAvg across heterogeneous domains (credit-risk + electricity forecasting) help, hurt, or leave per-domain performance unchanged? It produces asymmetric effects that cannot be summarised by one aggregate metric. Another honest negative result.

Fully-funded, salaried PhD positions in Europe in trustworthy AI, responsible AI, uncertainty quantification, physics-informed ML, or cross-domain evaluation methodology.

ORCID: 0009-0009-3177-8492