CV
Education
Thesis: Learning by Relaxation: Equilibrium Propagation for Neuromorphic Generative Modelling. Supervisors: Prof. Claudio Castelnovo (Cambridge), Dr Joachim Wabnig (Nokia Bell Labs). Thesis submission May 2027.
Master of Mathematical and Theoretical Physics, Distinction. BA Physics, First Class — ranked 1st in a cohort of ~200 in second-year examinations.
Experience
Proved that a predictive-coding network with one added linear energy term carries the gradient of an Energy Matching potential, and trained it entirely by Equilibrium Propagation to 3.64 FID on CIFAR-10 (3.54 for backprop), the first backprop-free generative model at this quality (PotentialEP). Introduced GradEP, extending EP from shaping energy minima to training energy gradients, and built FlowEqProp, the first EP-trained flow-matching generative model (Best Paper, ICONS '26). Developed EP training for Oscillator Ising Machines matching backprop on MNIST under realistic hardware constraints; a hardware implementation is under way with Nokia and TU Eindhoven. Uncovered a saddles-to-minima topological crossover governing glassy dynamics on rugged energy landscapes (submitted to Nature Communications).
Led short-horizon risk forecasting: gradient-boosted and interpretable EBM models over 1,500+ engineered features and 20M+ records, improving recall by 50% over prior baselines, plus LLM pipelines for semantic incident categorisation and case-note surfacing.
Evaluated frontier LLM abilities in model compression, mechanistic interpretability, and feature steering for ML R&D tasks.
Co-authored two Physical Review B papers on multi-band superconductivity and impurity effects in FeSe.
Enhanced the FETCH2 radiation modelling codebase; presented benchmarking results at Imperial College.
Teaching & supervision
Supervised undergraduate Statistical Mechanics and Mathematical Methods at Cambridge for three years (sole instructor, ~20 students). Supervised a Master’s dissertation on saddles-to-minima crossovers in kinetically constrained Ising models (2024–25); co-supervising a second from 2026.
Service
Reviewer, NeurIPS Workshop on Machine Learning and the Physical Sciences (ML4PS), 2025–2026.
Awards
- Best Paper Award, International Conference on Neuromorphic Systems (ICONS '26) — FlowEqProp
- Selected participant, ELLIS Cambridge Summer School on Probabilistic Machine Learning, 2026
- G-Research Grant (£2,000), 2025
- ICONS Student Travel Award ($1,000), 2025
- EPSRC CASE PhD Studentship with Nokia UK, 2022
- Winton Capital Prize, 1st of ~200 in Oxford Physics second-year examinations, 2020
Technical skills
Languages: Python, Julia, MATLAB, C++, Java, R, FORTRAN.
ML: PyTorch, JAX, NumPy, Hugging Face, XGBoost, LightGBM, InterpretML.
Infrastructure: SLURM/HPC, AWS, Git, Bash.
Full publication list on the publications page.