River AI
Software Engineer, GPU Kernels
$200K–$420K
Palo Alto, CA +1 · Full-time · Posted Sep 11, 2026
- Location
- Palo Alto, CA
Palo Alto, California, United States - Education
- Bachelor’s degree in Computer Science, Computer Engineering, or equivalent practical experience.
- Visa sponsorship
- We sponsor visas and are committed to supporting the process for the right candidate.
About the job
Build and optimize GPU kernels that accelerate River AI’s model training and inference infrastructure.
What you'll do
- Implement and optimize kernels for attention, matrix multiplication, expert routing, and low-precision computation.
- Profile real workloads, validate correctness, and integrate performance improvements into training and inference runtimes.
- Develop reproducible benchmarks and optimize fine-tuning and reinforcement learning operations.
What they ask for
- Bachelor’s degree in Computer Science, Computer Engineering, or equivalent practical experience.
- Experience optimizing GPU kernels using CUDA, Triton, CUTLASS, CuTe, or comparable tools.
- Strong understanding of GPU architecture, memory hierarchies, and parallel execution.
- Proficiency in C++ and Python; strong foundations in linear algebra, floating-point arithmetic, and numerical computing.
Nice to have
- Experience optimizing for NVIDIA Blackwell or Hopper GPUs, or working on attention, mixture-of-experts, grouped GEMMs, or low-rank adapters.
- Experience with backward passes, gradient validation, FP8/FP4, quantized weight layouts, or custom operators in PyTorch, SGLang, vLLM, or similar frameworks.
Pay
$200,000 - $420,000 USD per year (base pay)
Tools
CUDA, Triton, CUTLASS, CuTe, C++, Python, PyTorch, SGLang
AI focus
Develop compute kernels and primitives to accelerate AI model training, inference, fine-tuning, and reinforcement learning.
Model-written summary · the original posting is authoritative