Loading portfolio
I’ve built and contributed to a variety of systems, some for companies, some for curiosity. Each one pushed me to design better architectures, write cleaner code, and explore how intelligence fits into engineering.

Applied ML research at the Biocomputing and Development Systems Group, University of Limerick, on autoencoder-based anomaly detection for wind turbine fleets.
📈 Raised fleet-wide anomaly-detection AUC from 0.50 to 0.735, cutting the models needed from one per turbine to a single generalised model.
Website

An AI-powered marketing platform that automates creative production and testing to scale ad performance on Facebook and TikTok, processing 10M+ daily events at 99.9% uptime.
📈 Cut infrastructure costs 30% and churn 3% through automated reporting and system optimization.
Website

India's largest digital radiology network using AI to deliver faster, more accurate medical image reports. Also led backend development of the Order Management System for Tata 1MG's radiology integration.
📈 Reduced report load times 96% (10s → 0.36s) and cut radiology turnaround time 25% via an intelligent auto-assignment algorithm.
Website
MSc dissertation: a fifteen-experiment ablation study testing whether an anomaly detector trained on one wind turbine generalises well to unseen turbines.
Details

AI-assisted cocktail design: given mandatory ingredients and target flavors, custom ML models guide a genetic algorithm to evolve optimized recipes, deployed on AWS.
Details

RapidMock accelerates frontend development by mocking endpoints with real responses like 200, 404, 500 and many more. All with a single click and no change in frontend code.
rapidmock.com

Applied ConvNeXt-Tiny - a state of the art pure CNN architecture, to classify 300,000 hand-drawn Google QuickDraw sketches across 30 categories, achieving an accuracy of 87.88%
Github

A Genetic Programming (GP) Classifier with multi-objectie fitness function for the Adult Income Dataset that was amongst the top 4 with an accuracy of over 78%.
Github