I build optimization and machine-learning systems grounded in physics. At Atomic Industries I lead the team that turns customer CAD into manufacturable injection-mould tooling — computational geometry, thermal simulation, and large-scale optimization in one pipeline.
Systems I designed and built, from GPU geometry kernels to distributed optimizers and public research software.
I lead the team building the platform that takes customer CAD through to manufacturable injection-mould tooling. The internal work is not public; the short version is computational geometry, GPU compute and large-scale optimization applied to a hard physical design problem.
An automatically differentiable foreground extension to CMBLensing.jl, accelerated with CUDA.jl. Sparse approximations and preconditioners cut the cost of a log-likelihood evaluation by roughly 100×, making inference across tens of A100 nodes practical.
Coordinated a ten-plus person collaboration across Berkeley Lab, University of Washington, Paris-Saclay and ChaLearn to build uncertainty-aware machine-learning benchmarks for high-energy physics, served on the Perlmutter supercomputer.
The de facto standard simulation package for the microwave sky, with 300+ citations. Used by the Simons Observatory, CMB-S4 and the Atacama Cosmology Telescope; now maintained by the Pan-Experiment Galactic Science group.
Simulation, optimization and machine learning for physical systems — especially where the geometry is unforgiving and the loop is expensive.