Machine Learning Research Scientist
Generative machine learning and sampling algorithms — from the cosmos to the cell.
I'm a Senior Principal Scientist at Johnson & Johnson, working on machine learning for the life sciences.
Previously, I was a research scientist at SandboxAQ, where I led the release of SAIR — the largest publicly available dataset of protein–ligand structures — and worked on protein folding.
Before that, I worked on structure-based protein design at Dreamfold, and was a postdoctoral researcher at Mila, focused on GFlowNets and other generative ML methods.
My earlier career was in astrophysics. I received my Ph.D. from the University of Cambridge, co-led the accelerated forward modelling group at the Simons Collaboration in Learning the Universe, and contributed to the cosmological analyses of the Dark Energy Survey and the Planck Collaboration.
I'm especially interested in generative machine learning and sampling algorithms.
A few of the research efforts I've contributed to. If any of them spark a conversation, please get in touch.
Structurally Augmented IC50 Repository
The largest publicly available dataset of protein–ligand structures labelled with binding affinities. I led the release of SAIR at SandboxAQ to help the community train better models for structure-based drug discovery.
Read the announcementSampling algorithms for science
Research on GFlowNets and other generative ML methods for scientific discovery, spanning my postdoc at Mila and structure-based protein design at Dreamfold. Focused on diverse, high-quality sampling from complex energy landscapes.
Related publicationsAccelerated forward modelling
Co-leading the accelerated forward modelling group at the Simons Collaboration in Learning the Universe. Earlier work includes cosmological analyses for the Dark Energy Survey and the Planck Collaboration.
Collaboration siteFor collaborations, outreach, or questions about any of my work, reach out by email or find me on one of the platforms below.
