HPC GPU Software Fellow: Lead AI & Sci Computing (Hybrid, Austin)

Socket.dev; Tennessee; Hybrid; USD 200,000 - 320,000; Full time

Benefits offered by this job

Benefits at a glance

Job summary

AMD’s AI GPU Software (AGS) team seeks a High Performance and Scientific Computing Fellow. This IC role drives technical vision across HPC GPU libraries in the ROCm platform and contributes to multi-generation AMD GPUs.

The ideal candidate brings deep expertise in mathematical algorithms, GPU performance analysis, and distributed systems, with hands-on work in C/C++/Fortran, HIP, CUDA or OpenCL, MPI/SHMEM, and a proven record of performance optimizations.

Qualifications

• Strong background in mathematical algorithms for dense and sparse linear algebra.
• Experience with direct and iterative solvers and HPC libraries.
• Proven ability to design, implement, and optimize parallel algorithms on top-tier supercomputers.

Responsibilities

• Define technical vision and drive software strategy across ROCm HPC libraries.
• Lead performance analysis, tuning, and algorithmic innovation across HPC libraries.
• Provide technical mentorship to senior engineers and influence best practices.
• Communicate complex findings to senior leadership and external partners.
• Represent AMD in external forums, benchmarks, and customer engagements.
• Collaborate with AMD and partners to improve ROCm HPC and AI apps and tools.

Skills

Mathematics algorithms

Dense & sparse linear algebra

GPU performance analysis

C/C++/Fortran

HIP/CUDA/OpenCL

MPI/SHMEM

Distributed systems

Software architecture

Education

Bachelor's degree in CS/EE/Applied Math

Advanced degree preferred (MS/PhD)

Tools

HIP

CUDA

OpenCL

MPI

SHMEM

ROCm

Job description

AMD’s AI GPU Software (AGS) team seeks a High Performance and Scientific Computing Fellow. This IC role drives technical vision across HPC GPU libraries in the ROCm platform and contributes to multi-generation AMD GPUs.

The ideal candidate brings deep expertise in mathematical algorithms, GPU performance analysis, and distributed systems, with hands-on work in C/C++/Fortran, HIP, CUDA or OpenCL, MPI/SHMEM, and a proven record of performance optimizations.