Liquid AI
Member of Technical Staff - GPU Performance Engineer
San Francisco +2 · Hybrid · Staff · Full-time · Posted Jul 29, 2025
- Location
- San Francisco
Remote
Boston
About the job
Develop and ship custom GPU kernels, integrating research ideas into production pipelines and improving training and inference performance through profiling and benchmarking.
What you'll do
- Write and integrate high-performance GPU kernels into PyTorch pipelines, with correctness tests and numerics checks.
- Profile and optimize training and inference workflows; build benchmarks and regression guardrails.
- Collaborate with researchers to turn ideas into production performance improvements.
What they ask for
- Authored custom CUDA kernels, beyond calling cuDNN/cuBLAS.
- Strong understanding of GPU architecture and performance, including memory hierarchy, warps, shared memory/register pressure, and bandwidth-versus-compute limits.
- Proficiency with low-level profiling using Nsight Systems/Compute and performance methodology.
- Strong C/C++ skills.
Nice to have
- CUTLASS experience and tensor core utilization strategies.
- Triton kernel experience and/or PyTorch custom op integration; benchmark harnesses and performance regression tests.
Tools
CUDA, cuDNN, cuBLAS, Nsight Systems, Nsight Compute, C/C++, CUTLASS, Triton
AI focus
Optimize custom kernels and training, post-training, and inference workflows for the company's AI models; ship research-driven performance techniques.
Model-written summary · the original posting is authoritative