About

I build models that stand in for measurements too slow or too expensive to make: interatomic potentials for carbon, property prediction for small molecules, sequence models for antibody engineering. The through-line is computational models for real-world scientific problems, and getting them used by people who are not modellers.

I am currently a machine learning engineer at SandboxAQ, working on drug discovery: small-molecule property prediction, ML engineering and protein language modelling. Before that I was a research scientist in protein engineering at AbCellera in Vancouver, where I built AI and simulation-driven workflows for antibody selection and engineering, and led their deployment across more than thirty clinical targets.

Earlier, I was an industrial research fellow split between the University of Cambridge and Happy Electron, designing machine-learned interatomic potentials for energy materials. My PhD was in physics at UCL, on accuracy and transferability in machine learning potentials for carbon, which is the work that produced GAP-20. Before that, chemistry at Warwick, and a year at DSM in the Netherlands developing a 96% bio-renewable alternative to petrochemical alkyd coatings.

Away from the desk

In 2025 I took a career break to travel, climb, mountaineer and hike. I also learned web development and spent a good deal of time experimenting with language models. This site is downstream of both. Ask me about any of it.

Elsewhere

Google Scholar is the canonical list of publications, across physics, chemistry, machine learning and cancer research. Code is on GitHub, and there is a LinkedIn if that is more useful. The CV has the full record.

The best way to reach me is on LinkedIn.