Senior GPU Kernel Engineer for AI Infrastructure

Designworks Talent LLC; Bellevue (KY); Hybrid; USD 170,000 - 250,000; Full time

Benefits offered by this job

Hybrid work model

Medical, dental, vision insurance

Job summary

Designworks Talent LLC is seeking Senior GPU Performance / Kernel Engineers to optimize the data plane powering large-scale AI workloads. This hybrid Bellevue-based role focuses on maximizing GPU utilization, reducing latency, and throughput by tuning kernels and eliminating bottlenecks.

You will work across training and inference environments, collaborating with AI infrastructure, ML, and platform teams to understand workload characteristics and drive efficient GPU performance.

Qualifications

• Proficient in GPU kernel development and performance tuning.
• Experience optimizing GPU workloads for AI training/inference.
• Strong knowledge of parallel computing and memory hierarchies.

Responsibilities

• Profile and optimize GPU kernels to reduce latency and increase throughput.
• Identify GPU data-plane bottlenecks across large-scale workloads.
• Tune workloads for training and inference in AI infrastructure.
• Collaborate with AI infra, ML, and platform eng teams to optimize system behavior.
• Develop benchmarking methods and performance metrics for GPU fleets.
• Evaluate new GPU tech and performance tools as hardware evolves.
• Contribute to practices that boost GPU efficiency and reliability.

Skills

GPU kernel development

Performance optimization

Parallel computing

Systems programming

Tools

Nsight Systems

Nsight Compute

ROCm profiling tools

Job description

Designworks Talent LLC is seeking Senior GPU Performance / Kernel Engineers to optimize the data plane powering large-scale AI workloads. This hybrid Bellevue-based role focuses on maximizing GPU utilization, reducing latency, and throughput by tuning kernels and eliminating bottlenecks.

You will work across training and inference environments, collaborating with AI infrastructure, ML, and platform teams to understand workload characteristics and drive efficient GPU performance.