Kate Clark works at the interface of applications, algorithms, and parallel computation. At NVIDIA, her work focuses on mixed-precision solvers, high-precision GEMM emulation, and HPC architecture co-design—ensuring that tomorrow’s scientific workloads run optimally on next-generation GPU architectures. Her work emphasizes harnessing the full memory bandwidth and hierarchy of evolving GPU systems to unlock performance at scale. She is the lead of QUDA, a widely used GPU-accelerated library for Lattice QCD.

Before joining NVIDIA, she was a researcher at Harvard University working on real-time radio astronomy signal processing, where she developed GPU-accelerated cross-correlation pipelines. She subsequently held a postdoctoral position at Boston University, focusing on multigrid solver algorithms. She received her PhD from the University of Edinburgh, Scotland, where her doctoral research centered on Monte Carlo algorithms for Lattice QCD.