Relace
GPU Kernel Engineer
San Francisco · On-site · Full-time · Posted Oct 21, 2026
- Experience
- Five or more years of low-level performance work; examples include operating systems, drivers, databases, compilers, trading systems, and inference engines.
About the job
Write and tune low-level GPU code to increase inference throughput, reduce model running costs, and keep the GPU fleet busy.
What you'll do
- Optimize GPU code for performance and cost.
- Shrink models without reducing their quality.
- Identify and address causes of GPU idle time.
What they ask for
- Strong C or C++ and low-level knowledge of memory, concurrency, caches, and hardware behavior.
- Track record of significantly improving performance and explaining the bottleneck and changes.
- Measure and profile performance before and after changes, using reliable test setups.
Nice to have
- Operating system, hypervisor, driver, firmware, or real-time work; performance tools; compiler work.
- GPU experience in industry or open source.
Tools
C, C++, GPU, CUDA, Kubernetes
AI focus
Optimize GPU inference workloads and shrink models to reduce serving costs without reducing model quality.
Model-written summary · the original posting is authoritative