Data scientist with 20+ years of experience modelling complex, noisy systems — originally in astrophysics, now applied to real-world data problems.

My work focuses on machine learning, statistical inference, and time series analysis, with an emphasis on extracting signal from difficult datasets and supporting decision-making under uncertainty.

• Fraud Detection (Machine Learning)

  Developed supervised models to identify rare events in highly imbalanced data, achieving strong precision while maintaining a low alert rate. Focused on threshold optimisation, uncertainty, and real-world trade-offs between detection and operational cost.

• A/B Testing Toolkit (Statistical Inference)

  Built reusable Python tools for comparing groups using confidence intervals and hypothesis testing. Designed to support practical decision-making in experimentation workflows.

• Time Series Forecasting Toolkit

  Created an interactive framework for comparing forecasting models (ARIMA, Holt-Winters, Prophet) with built-in backtesting and error analysis to evaluate real-world performance.

• Deep Learning for Regression

  Implemented neural network models for continuous parameter estimation from high-dimensional data, including full pipelines for preprocessing, training, validation, and uncertainty assessment.

Python (pandas, NumPy, scikit-learn, TensorFlow), statistical modelling, machine learning, time series forecasting, hypothesis testing, data visualisation.