Low-Latency GPU Kernel Engineer

SIG Susquehanna; Bala Cynwyd (PA); On-site; USD 120,000 - 160,000; Full time

Job summary

SIG Susquehanna in Bala Cynwyd, Pennsylvania, is seeking a GPU Performance Engineer to develop optimized CUDA kernels for low-latency inference. The ideal candidate has strong CUDA programming skills and expertise in GPU architecture. Responsibilities include optimizing kernel performance, analyzing models for computational bottlenecks, and enhancing production inference systems. A PhD in a relevant field is preferred. Join an engaging environment committed to performance improvements and innovation in computational efficiency.

Qualifications

• Strong proficiency in writing and optimizing CUDA kernels.
• Solid programming experience in C/C++.
• Deep understanding of GPU architecture and performance tradeoffs.
• Ability to reason about numerical stability and precision.
• Strong problem-solving skills.

Responsibilities

• Design, implement, and optimize custom CUDA kernels.
• Develop GPU implementations tailored to specific model structures.
• Analyze computational bottlenecks and identify optimization opportunities.
• Collaborate with researchers to translate models into high-performance pipelines.
• Improve end-to-end inference performance.

Skills

CUDA optimization

C/C++ programming

GPU architecture understanding

Problem-solving in low-level systems

Education

PhD in mathematics, physics, computer science, engineering, or related field

Tools

ONNX Runtime

TensorRT

Triton

TVM

Job description

SIG Susquehanna in Bala Cynwyd, Pennsylvania, is seeking a GPU Performance Engineer to develop optimized CUDA kernels for low-latency inference. The ideal candidate has strong CUDA programming skills and expertise in GPU architecture. Responsibilities include optimizing kernel performance, analyzing models for computational bottlenecks, and enhancing production inference systems. A PhD in a relevant field is preferred. Join an engaging environment committed to performance improvements and innovation in computational efficiency.