Presenter

Jeremiah Wilke is a software engineer in the CUDA software group at NVIDIA. He received his PhD in computational chemistry from the University of Georgia before joining the scalable computing and analysis group at Sandia National Labs in Livermore, CA. In his 10 years at Sandia, Jeremiah worked on a variety of problems including distributed programming models, scalable network architectures, exascale performance modeling through discrete event simulation and skeleton applications (SST), and performance portability in C++ (Kokkos). More recently, Jeremiah worked on compiler autotuning and fleetwide performance monitoring in the Core ML team at Google before joining NVIDIA to work on the Legate distributed runtime. His most recent work has focused on Jax and combining ML compiler stacks with task-based runtimes (Realm) to enable MPMD DL workflows.