CUDA Kernel Engineer (Remote US)
Pragmatike17 days ago
Cambridge, MA, USA +5 moreMid Level / Senior
Responsibilities
- Design, implement, and optimize custom CUDA kernels for NVIDIA GPUs.
- Profile GPU workloads using tools like Nsight Compute and nvprof.
- Analyze and eliminate performance bottlenecks in GPU workloads.
- Improve GPU memory pipelines and ensure proper memory coalescing.
- Collaborate with AI systems and backend distributed systems teams.
- Contribute to GPU architecture decisions and performance-engineering best practices.
Requirements
- Proven experience building NVIDIA CUDA kernels from scratch.
- Strong optimization skills for kernels, including tiling strategies and warp scheduling.
- Deep understanding of CUDA threads, warps, and GPU memory hierarchy.
- Experience diagnosing PCIe bottlenecks and optimizing host-device transfers.
- Familiarity with C++, CUDA runtime APIs, and GPU debugging tools.
Benefits
- Competitive salary & equity options
- Sign-on bonus
- Health, Dental, and Vision insurance
- 401k plan