Remote CUDA Kernel Engineer – Optimize GPU Performance

Pragmatike; Cambridge (MA); On-site; USD 150,000 - 230,000; Full time

Benefits offered by this job

Salary + equity

Sign-on bonus

Health/Dental/Vision

401k

Job summary

Pragmatike, a fast-growing AI startup recognized as a Top 10 GenAI company by GTM Capital and founded by MIT CSAIL researchers, offers a remote US role for an expert CUDA/kernel engineer. The position ASAP requires English and focuses on designing, implementing, and optimizing custom CUDA kernels for NVIDIA GPUs while profiling and debugging performance.

You'll collaborate with AI systems and backend teams to push production-grade performance.

Qualifications

• Proven track record building NVIDIA CUDA kernels from scratch.
• Strong ability to optimize kernels (tiling strategies, occupancy tuning, shared memory design).
• Deep understanding of CUDA threads, warps, blocks, grids, memory hierarchy and coalescing, warp divergence.

Responsibilities

• Design, implement, and optimize CUDA kernels for NVIDIA GPUs, maximizing occupancy and memory throughput.
• Profile GPU workloads with Nsight Compute/Systems, nvprof, and CUDA-MEMCHECK.
• Identify and fix bottlenecks: warp divergence, uncoalesced memory access, register pressure, PCIe transfers.
• Improve memory pipelines (global/shared/L2/texture) and ensure proper memory coalescing.
• Collaborate with AI systems, model acceleration, and backend distributed teams.
• Contribute to GPU architecture decisions and internal performance practices.

Skills

CUDA kernel design

Kernel optimization

GPU architecture understanding

C++ programming

Profiling & debugging

Tools

Nsight Compute

Nsight Systems

nvprof

CUDA-MEMCHECK

PTX/SASS analysis

Job description

Pragmatike, a fast-growing AI startup recognized as a Top 10 GenAI company by GTM Capital and founded by MIT CSAIL researchers, offers a remote US role for an expert CUDA/kernel engineer. The position ASAP requires English and focuses on designing, implementing, and optimizing custom CUDA kernels for NVIDIA GPUs while profiling and debugging performance.

You'll collaborate with AI systems and backend teams to push production-grade performance.