Senior GPU Performance Engineer - AI Workloads
Core
Optimize AI workloads (transformers, diffusion models) on large-scale GPU systems by performing deep kernel-level analysis and low-level optimizations.
Role type
Senior IC GPU Performance Engineer (AI/Hardware)
Builds
Industry-leading AI performance on AMD Instinct GPUs
Domain
Semiconductor / AI Infrastructure
Deliverable
production ML models
Required skills
GPU kernel-level performance analysis, CUDA/HIP/ROCm, GPU microarchitecture, C++, Python, performance counters, ISA/assembly analysis, microbenchmarking, Linux, Docker, Git, AI frameworks (PyTorch, JAX, Megatron-LM), distributed computing (NCCL, MPI)
Preferred skills
Hardware/software co-design, cross-functional collaboration on kernel/compiler/runtime teams, translating workload gaps to microarchitecture improvements
Technologies
CUDA, HIP, ROCm, PyTorch, JAX, Megatron-LM, NCCL, RCCL, MPI, Docker, Linux
Responsibilities
Profile and optimize large-scale AI workloads across multi-node systems; Identify bottlenecks in compute pipelines, memory hierarchy, and instruction execution; Root-cause performance gaps using profilers and assembly analysis; Collaborate with compiler and distributed systems teams to improve efficiency; Develop benchmarks and microbenchmarks for foundation models; Provide insights to architecture teams for future GPU designs; Present findings to leadership to shape roadmaps
Seniority
Senior, hands-on IC
