Projects
ML engineering, platform infrastructure, and analytics projects — from enterprise GenAI evaluation to Bayesian inference frameworks.
The Invincibles Decoded
Arsenal 2003/04 Football Analytics
An interactive analytics dashboard exploring Arsenal's historic unbeaten 2003/04 Premier League season. Built with StatsBomb open data, featuring shot maps, passing networks, and player heatmaps across all 38 matches.
- ▸D3.js visualisations: xG shot maps, passing networks, and binned heatmaps
- ▸StatsBomb open data pipeline processing 38 matches, 544 shots, 20K+ passes
- ▸Interactive match and player filtering across all dashboard panels
Drafty
FPL Draft League Analytics
A live analytics dashboard for FPL draft leagues. Tracks bracket standings, league position over time, transfer history, and bench efficiency with automated weekly data refreshes.
- ▸D3.js visualisations: standings timeline, bench efficiency bars, transfer summaries
- ▸Python/DuckDB pipeline with automated GitHub Actions weekly refresh
- ▸Next.js static export deployed on Vercel with zero server costs
BayesLands
Bayesian Landscape Evolution
Bayesian inference framework using Parallel Tempering MCMC for Badlands — a landscape evolution model. Estimates geological parameters with surrogate-assisted sampling for computationally expensive simulations.
- ▸Published in Computers & Geosciences (Elsevier)
- ▸Surrogate neural networks to accelerate MCMC convergence
- ▸Parallel tempering across HPC clusters for multi-modal posteriors
BayesReef
Bayesian Coral Reef Modelling
A Bayesian inference framework for pyReef-Core using Metropolis-Hastings MCMC. Quantifies uncertainty in coral reef growth predictions by estimating parameters as full probability distributions.
- ▸Uncertainty quantification for reef growth under climate scenarios
- ▸Metropolis-Hastings sampler with adaptive proposal tuning
- ▸Integrated with pyReef-Core stratigraphic forward model