New York City Metropolitan Area
High Performance Computing Professional with 8+ years of experience working on Supercomputing technology
https://github.com/thomas-huber I am a Software Engineer on the RCCL team. I work on: - GPU Networking - Communications Middleware - RCCL/NCCL, Libfabric, MPI - Debugging and solving problems at large scale - Using agentic AI to accelerate engineering workflow
Developed new features and improved performance of the OPX libfabric provider. This low-level networking abstraction works in conjunction with Cornelis Network's Interconnect (NIC) products. Topics I've worked on: - NVIDIA GPU Programming (NCCL, CUDA, GDRCopy, GPUDirectRDMA, cudaIPC, P2P) - AMD GPU Programming (ROCm, HSA) - MPI - OpenSHMEM - Improving GPU offloading performance internode and intranode
https://crpl.cis.udel.edu/
https://crpl.cis.udel.edu/
REU Summer Intern funded by National Science Foundation to work on Indiana University’s Jetstream Supercomputer. Evaluated the performance of Indiana University's 'Jetstream' supercomputing system using the SPECOMP benchmark suite. By gathering statistics of a single node's performance across various compute-intensive tests in a controlled manner, it was possible to gain insight into the native performance of hardware vs performance of the hardware when running KVM hypervisor in production Attended PEARC2018 advanced computing conference as a student volunteer, presented research in poster competition and participated in student modeling challenge.