Our work spans diffusion, flow matching, and neural samplers — from mathematical foundations to what they make possible in science: protein design, molecular dynamics, and sampling from unnormalized densities.
The mathematics of modern generative modelling — flow matching, diffusion, and flow maps, from Riemannian manifolds to discrete spaces.
Generative models for scientific discovery: protein design, molecular dynamics, crystal structure prediction, and beyond.
Sampling from unnormalized densities, iterated energy matching, and amortized inference.
Adversarial machine learning and red-teaming of large language models.
This is the lab’s new home for technical writing — paper deep-dives, tutorials, negative results worth sharing, and the folklore that never fits in an eight-page paper. Every la...
We are always looking for exceptional PhD students, MSc students, and interns interested in generative models, geometry, and AI for science. Imperial PhD applicants: apply through the Department of Computing and mention Joey Bose as your preferred supervisor. Please read our representative papers before reaching out.
Join the lab