I am a Principal Applied Scientist / Principal Data Scientist with 15+ years of experience delivering production machine learning solutions at scale across advertising, media, music rights and finance. My work has included low-latency bidding systems, deep probabilistic models, click and conversion prediction, entity matching, audience targeting, recommender systems, sentiment analysis and active learning.
I have held senior data science, applied science and engineering roles at The Trade Desk, Kobalt Music, Schibsted, Quantcast, Struq, Morgan Stanley and Dublin City University.Â
Most recently, I owned the Predictive Clearing product, developing deep probabilistic models that scored billions of auction bids in real time, helped protect advertisers from overpaying in first-price auctions, and contributed to billions of dollars in advertiser savings. I also helped shape foundation-model work using multi-task learning to better utilise data and compute with Ray.
My career is focused on applying machine learning at scale to automate decisions, improve performance and create measurable business impact. My technical interests include large-scale deep learning, probabilistic modelling, NLP, active learning, recommender systems and production ML infrastructure.
Ph.D. in Computer Science from Trinity College Dublin in 2008, in the area of Artificial Intelligence. My research was funded through the Embark Initiative scholarship from the Irish Research Council for Science Engineering and Technology.
M.Sc. in Computer Science (Networks and Distributed Systems) from Trinity College Dublin
B.Sc. (Hons) in Computer Applications (Software Engineering) from Dublin City University
My thesis investigated history-based query selection strategies for Active Learning, contributing novel and efficient query selection algorithms that leveraged information from classification predictions in previous iterations. Other directions included unsupervised dimensionality reduction techniques for text-based domains in an active learning setting, and a reinforcement learning technique to dynamically switch among alternative query selection strategies based on their relative performance.
Large-scale Deep Learning
Deep Probabilistic Modelling
Natural Language Processing (NLP)
Traditional Machine Learning
Recommender Systems
Active Learning
Sentiment Analysis