Member of Technical Staff, GPU Kernels
Core
Hand-optimizing GPU kernels for pre-training, post-training, and inference workloads to establish performance ceilings before compiler search.
Role type
Senior IC GPU Kernel Engineer
Builds
High-performance GPU kernels for NVIDIA, AMD, TPU, and Trainium hardware
Domain
High-performance computing, GPU architecture, compiler optimization
Deliverable
production ML models
Required skills
CUDA, C++, PTX, SASS, Nsight Compute, Nsight Systems, low-level profiling, microarchitectural analysis, formal verification
Preferred skills
LLVM/MLIR, superoptimization, program synthesis, distributed AI training, high-speed interconnects, HPC, driver development
Technologies
CUDA, ROCm/HIP, PTX, SASS, cubins, Nsight Compute, Nsight Systems, LLVM, MLIR, NVLink, NVSwitch, InfiniBand, RoCE
Responsibilities
Write and hand-optimize kernels for real workloads; profile at the microarchitectural level; debug clock behavior and thermal throttling; convert hard-won knowledge into machine-searchable structures; work below PTX at the ISA-level; build performance models and microbenchmarks; collaborate with the formal correctness team.
Seniority
Senior, hands-on IC
