About
I work on the physics of machine learning: training algorithms that let physical systems learn by settling, rather than by backpropagation. My PhD, joint between the Math & Algorithms Group at Nokia Bell Labs and the Theoretical Condensed Matter Group at the University of Cambridge, centres on Equilibrium Propagation — from state-of-the-art EP training on analog Oscillator Ising Machines, now being implemented in hardware with Nokia and TU Eindhoven, to GradEP, which extends EP to train energy gradients and enabled the first EP-trained flow matching generative model (Best Paper, ICONS ‘26).
A second thread connects this to statistical physics: gradient descent on rugged energy landscapes, where we uncovered a saddles-to-minima topological crossover controlling glassy dynamics, using the Rubik’s Cube as a model system. Alongside the PhD, I lead machine learning work on violence risk forecasting at the Alan Turing Institute & UK Ministry of Justice.
News
- FlowEqProp wins Best Paper at ICONS ‘26 Award
- Selected for the ELLIS Cambridge Summer School on Probabilistic Machine Learning
- Presented Learning at the Speed of Physics at the NeurIPS ML4PS workshop
- Started as ML Technical Project Lead, Alan Turing Institute & UK Ministry of Justice
- Rubik’s Cube glassiness paper submitted to Nature Physics
Selected publications
- FlowEqProp: Training Flow Matching Generative Models with Gradient Equilibrium PropagationIn Proceedings of the International Conference on Neuromorphic Systems (ICONS ’26), 2026