About this opportunity
Confidential lists this GPU Kernel Developer for AI Model Evaluation opportunity in Tarrs, Pennsylvania. Review the employer’s description below for duties, qualifications and application requirements.
Job description
Role Overview
Assess GPU and accelerator kernel development tasks that support the training and evaluation of advanced AI models. You will review task quality, numerical correctness, completeness, benchmarking fairness, scope, and compilation and runtime validity across a range of kernel-development scenarios, then deliver clear written feedback using defined evaluation rubrics.
Key Responsibilities
Evaluate GPU and accelerator kernel tasks for quality, correctness, completeness, and appropriate scope.
Review numerical-validation approaches, performance-benchmarking fairness, and compilation and runtime validity.
Provide precise, rubric-based written feedback across diverse kernel task types.
Qualifications
3+ years of hands-on experience developing, optimizing, or verifying GPU or accelerator kernels in at least two of CUDA, Triton, NKI, or Pallas for JAX.
Strong knowledge of kernel numerical-correctness standards, including absolute, relative, and ULP tolerances and selection of reference implementations.
Experience profiling and benchmarking performance with tools such as Nsight, NCU, roofline analysis, or framework-native profilers.
Familiarity with compilation and runtime failure modes, including driver mismatches, out-of-memory errors, launch-configuration errors, shape or stride mismatches, and autotuning failures.
Experience with at least three kernel task types: specification-based generation, cross-framework translation or lowering, hardware-target migration, debugging, performance optimization, or operator fusion.
Preferred Qualifications
Experience in both NVIDIA GPU environments, including CUDA or Triton, and custom accelerator environments, including NKI, Pallas, or TPU.
Background in compiler engineering, MLIR, or intermediate-representation lowering.
Knowledge of memory-hierarchy optimization, including shared-memory tiling, register pressure, bank conflicts, and coalescing patterns.
Contributions to kernel libraries such as cuBLAS, cuDNN, Triton community kernels, or JAX/XLA custom calls.
Work Terms
Remote role open to candidates located in the United States.
Hourly engagement.
Compensation
$70 to $90 per hour.
Who can apply
Review the original listing for work authorization, qualifications and employer requirements.