GPU Kernel Engineer

Coda Robotics; San Francisco (CA); On-site; USD 100,000 - 120,000; Full time

Job summary

Coda Robotics is looking for an experienced engineer to join their founding team, focusing on low-level compute kernels to enhance robotic foundation models. The ideal candidate will have substantial experience in systems programming (C/C++, assembly), expertise in GPU optimizations, and familiarity with ML framework internals. Responsibilities include leading engineering teams, integrating kernel optimizations, and pioneering high-velocity development culture. Compensation includes a salary range of $100,000 - $120,000 plus equity and benefits.

Qualifications

• Proven experience in low-level systems programming (C/C++, assembly) targeting CPU and GPU architectures.
• Expertise in developing and optimizing compute kernels (CUDA, ROCm, OpenCL, SIMD intrinsics).
• Deep understanding of performance profiling tools (nvprof, perf, Intel VTune).

Responsibilities

• Lead a team of kernel and system engineers focused on performance-critical code.
• Design, implement, and optimize custom compute kernels for CPU and GPU.
• Integrate kernel optimizations into distributed ML frameworks (e.g., PyTorch, TensorFlow).

Skills

Low-level systems programming

CUDA

C/C++

Memory optimization

GPU architectures

Performance profiling tools

Job description

Coda Robotics is scaling the compute infrastructure that powers next‑generation robotic foundation models. As training and inference workloads grow, we need kernel‑level innovations to reduce latency, memory usage, and energy consumption. You will join Coda's founding team to architect and optimize low‑level compute kernels, drivers, and runtime components—making model training and inference significantly cheaper and faster.

Responsibilities

• Lead a team of kernel and system engineers focused on performance-critical code
• Design, implement, and optimize custom compute kernels for CPU (AVX/ARM NEON), GPU (CUDA/ROCm), and hardware accelerators
• Find bottlenecks in memory hierarchy, thread scheduling, and data movement
• Integrate kernel optimizations into distributed ML frameworks (e.g., PyTorch, TensorFlow) and orchestrate deployment in cloud and edge environments
• Explore OS and driver‑level enhancements—such as zero‑copy I/O, custom scheduling, and power management—to further boost throughput
• Define and own the technical roadmap for kernel and runtime subsystems, balancing performance, maintainability, and portability
• Drive rapid iteration, testing, and benchmarking cycles to validate improvements and de‑risk large‑scale rollouts
• Champion a high‑velocity culture that values bold technical ambition, clear accountability, and measurable impact

Requirements

• Proven experience in low‑level systems programming (C/C++, assembly) targeting CPU and GPU architectures
• Expertise in developing and optimizing compute kernels (CUDA, ROCm, OpenCL, SIMD intrinsics)
• Deep understanding of performance profiling tools (nvprof, perf, Intel VTune) and techniques for memory and compute optimization
• Strong familiarity with ML framework internals (PyTorch, TensorFlow) and integration of custom operations
• Experience with compiler design or code generation (LLVM, MLIR) is a plus

Compensation

Base salary range: $100,000 - $120,000 per year, plus strong equity and benefits