GPU Distributed System Researcher

Advanced Micro Devices, Inc•Bellevue, WA

11d•Hybrid

About The Position

AMD Research and Advanced Development is seeking a researcher to invent and implement novel programming models and runtime environments for high performance computing systems comprised of heterogeneous processors, accelerators, and scale-up/scale-out networks. The ideal candidate loves inventing and implementing new features, APIs, and abstractions to increase efficiency and optimize performance, bringing new systems to life and developing software that showcases the full power of novel hardware architectures. This role will help define the software stack of AMD’s future hardware accelerators.

Requirements

Must have experience in developing and debugging GPU and/or multi-threaded CPU applications.
Distributed system experience is desired.
Prior exposure to ML/HPC/datacenter software, middleware and device driver development, and familiarity with ML frameworks, OpenSHMEM, or MPI programming models are preferred.
The position involves investigating networking architectures with an emphasis on optimizing cluster-scale applications that are bounded by memory and/or inter-GPU communication performance.
Demonstrated experience in ideation, evaluation, and optimization of a research project.
Excellent written and oral communication skills, ability to organize and present complex technical information.
Strong analytical skills.
Strong skills in C/C++, Python and/or GPU programming.
Use of networking simulation platforms (NS-3, OMNeT++, etc.).
Experience with High Performance Computing networking.
Experience with the AMD ROCm software stack.
Version control systems such as Git.

Responsibilities

Invent, design and implement novel programming models and runtime environments for high performance computing systems comprised of heterogeneous processors and accelerators.
Implement innovative software solutions and demonstrate their effectiveness on prototype hardware.
Work on software enhancements that aim to potentially improve programmer productivity.
Drive the results of the research projects into product roadmaps.
Collaborate with teams within AMD Research and Advanced Development, product groups, and vendor partners to help improve AMD’s ML/HPC ecosystem.
Submit patentable inventions.
Clearly communicate research findings to academic, product and commercial audiences.