About
DPhil (PhD) student in Computer Science at the University of Oxford advised by Prof. Alfonso Bueno-Orovio. Previously research intern at UIUC Blender Lab (energy-based models). MSc AI (Distinction) from the University of Essex, advised by Prof. Luca Citi.
Research Interests
What are the fundamental invariances that can compose and create intelligence? My ultimate goal is to understand and build intelligence.
Many interesting scientific questions need modeling non-linear and chaotic dynamical systems. Some are intrinsically ill-posed and have non-unique solutions and need a probabilistic language. I'm interested in developing models able to represent such dynamics natively.
Inverse problems usually need to be solved from noisy and partially observed data. The question is how can we learn the global representation of the underlying dynamics under data constraints? Given that I tend to avoid handcrafting physics and learning-competing inductive biases explicitly (for generalizability and scalability reasons), data-efficient self-supervised learning methods are another line of work I'm keen to work on.
I've built TorchEBM that looks at these problems through the lens of statistical mechanics, differential and Riemannian geometry, and Optimal Transport.
News
- New Preprint Mixed-Field Matching: Time-Conditioned Transport with Energy-Based Refinement - Oct. 2026
- New Preprint How to Train Your Energy-Based Transformer: Understanding Stability in EBMs - Oct. 2026
- New Awarded EPSRC DPhil studentship at University of Oxford - Feb. 2026
- New CS DPhil offer at University of Oxford - Feb. 2026
- Joined CoSTAR National Lab as ML Research Engineer - Feb. 2026
- Research Collaboration with UIUC Blender Lab - Jan. 2025
- Full scholarship for MSc AI, University of Essex - Oct. 2023
Selected Publications
2026
Preprint
Mixed-Field Matching: Time-Conditioned Transport with Energy-Based Refinement
2026
Preprint
How to Train Your Energy-Based Transformer: Understanding Stability in EBMs
Selected Software
TorchEBM 🍓
Updated Official Website Read more
PyTorch library for energy-based models, diffusion, and flow matching. Implements samplers, score/flow/contrastive objectives, SDE/ODE integrators, interpolants, and mixed-precision training.
PyTorch · CUDA
WorldKernels 
High-throughput world model inference engine. Supports BiD and AR world models (DreamDojo, Cosmos, DreamZero, etc.) with persistent KV-cache, CUDA graph capture, continuous batching, speculative decoding, and torch.compile fusion. Sub-millisecond scheduling via async token queues.
PyTorch · CUDA · Python
cuRBLAS 🍒
GPU-accelerated randomized linear algebra. CUDA kernels for Hutchinson trace estimation, randomized SVD, and probabilistic matrix operations, useful for sliced score matching and large-scale ML.
C++ · CUDA · Python
