Project Page
gpCAM is an API and software for Uncertainty Quantification, Bayesian Optimization, and Autonomous Data Acquisition. It is entirely data-source agnostic and can therefore be used to steer simulations and experiments. The tool is based on a flexible Gaussian process (GP) engine. gpCAM (through fvgp) holds the world record for exact scalable GPs. It is targeted at users who want easy and full control of customization options such as kernel, prior mean, noise, and acquisition functions. Visit the gpCAM website for more information.
fvGP is the Gaussian process engine running within gpCAM. The tool's main purpose it to provide a highly-customizable and HPC-ready implementation of Gaussian processes.
HGDL is an asynchronous distributed optimizer that is tailored to be used on high-performance computing architectures. It is based on a hybrid local/global optimization plus deflation which allows us to remove optima from the function.
Solutions to Eikonal and Wave Equations
Much of my work as a PhD candidate was concerned with developing algorithms and software that simulates wave propagation. The focus was on GPU-enabled finite-difference algorithms.