Presenter

Madan K. Sharma Timalsina is a NESAP-Postdoc at NERSC/LBNL, working on GPU-accelerated scientific workflows and scalable Python for high-energy physics. Recent efforts with the DUNE ND-LAr team optimize the larnd-sim pipeline on NERSC Perlmutter (AMD Milan–A100) and TACC Grace Hopper (ARM64–GH200) using Numba/CuPy and Nsight Systems/Compute, delivering up to 5X kernel speedups, ~32% end-to-end runtime reductions, and >50% peak-memory savings. His work extends to containerized, fault-tolerant pipelines with checkpoint-restart (DMTCP) across Docker, Shifter, Podman, and Apptainer, plus data-production frameworks (e.g., LZ Prompt Processing). He regularly mentors GPU optimization efforts and contributes to DOE projects, including DUNE, LZ Dark Matter experiment, and US CMS.