Summary
Data scientist with 5 years of experience building machine learning models that run in production. Shipped a fraud model that cut losses by £3M a year and a recommendation system that grew basket size by 8%. Experienced with Python, PyTorch, SQL and MLOps, and skilled at explaining results to executives.
Experience
- Built and deployed a gradient-boosted fraud model (AUC 0.93), reducing fraud losses by £3M a year
- Designed 25 A/B experiments with a Bayesian framework now standard across the company
- Cut model retraining time from 6 hours to 40 minutes by moving pipelines to MLflow and Spark
- Launched a product recommendation model that increased average basket size by 8%
- Reduced demand-forecast error by 22% using gradient boosting and weather features
- Mentored 3 graduate data scientists through their first production models