Remote Senior GPU Kernel Evaluator & Optimisation Expert

24-Mag Llc; New York, Northern (NY, KY); Hybrid; USD 83,000 - 110,000; Part time

Job summary

24-MAG LLC offers a specialised part-time consulting opportunity for experienced GPU and accelerator engineers with hands-on expertise in kernel development, numerical validation, performance optimisation, profiling, and low-level compute frameworks. This role focuses on evaluating GPU and accelerator kernel-development tasks for correctness, completeness, reproducibility, and performance quality.

You will review kernel implementations, benchmark methodology, compilation and runtime behaviour,

Qualifications

• 3+ years hands-on experience developing, optimising, or verifying GPU/accelerator kernels.
• Experience with at least two of CUDA, Triton, NKI, or Pallas (JAX).
• Strong numerical correctness concepts including absolute, relative, and ULP tolerances.
• Proven performance profiling, benchmarking, and debugging skills.

Responsibilities

• Evaluate GPU and accelerator kernels for technical correctness.
• Review kernel implementations from specifications or reference operators.
• Assess numerical behaviour and tolerances.
• Benchmark methodology and runtime behaviour review.
• Evaluate memory tiling, coalescing, and efficiency.
• Identify bottlenecks and ensure numerical correctness during optimisations.

Skills

GPU kernel development

Numerical correctness

Performance profiling

Debugging kernels

Technical feedback

Tools

Nsight

Nsight Compute

roofline analysis

CUDA

Triton

NKI

Pallas (JAX)

Job description

24-MAG LLC offers a specialised part-time consulting opportunity for experienced GPU and accelerator engineers with hands-on expertise in kernel development, numerical validation, performance optimisation, profiling, and low-level compute frameworks. This role focuses on evaluating GPU and accelerator kernel-development tasks for correctness, completeness, reproducibility, and performance quality.

You will review kernel implementations, benchmark methodology, compilation and runtime behaviour,