OpenRecently verified

GPU Kernel Expert

  • Software Engineering
  • Platform: Mercor
  • Location: United States
  • Level: Intermediate

$70-90/hrAs published by the platform.

Not sure this job fits you?

  • Add your resume and see every open Software Engineering job scored against your background. Free.
  • Your profile is never sent to the platforms. You decide where to apply.
Add your resume

About this job

This job involves evaluating the quality and correctness of GPU kernel and accelerator code developed to train and test AI models. You will provide structured feedback on numerical correctness, the validity of performance benchmarks, task definition, and viability for compilation and execution.

What you'll do

  • Assess the numerical correctness of GPU kernels and accelerators according to defined tolerance criteria
  • Analyze the fairness and validity of performance benchmarks using profiling tools
  • Verify the validity of tasks against specifications and identify compilation or execution errors
  • Provide structured written feedback based on an evaluation checklist

Requirements

CUDATritonNKIPallas (JAX)Performance profilingBenchmarkingCompiler engineeringMLIR
  • 3+ years of experience

Pay

$70-90/hr

Pay as published by the platform. It is not a guarantee of income or hours.

Location

Open to residents of: United States.

How to apply

You apply on work.mercor.com: create a profile, complete a skills assessment, then get matched to projects that fit your background.

All Mercor jobs and how the platform works

Source

Official posting: https://work.mercor.com/jobs/list_AAABoDrLqkDg31vugrpECa5L/gpu-kernel-expert

Last checked on November 20, 2026.

Similar jobs

  • Software Engineering
  • Platform: Alignerr
  • Level: Intermediate

Design and build high-performance C++ systems for AI data pipelines, evaluation workflows, and internal tooling. This role involves optimizing production codebases, developing backend services for data annotation and quality control, and collaborating with research teams to support model training infrastructure.

  • Software Engineering
  • Platform: Alignerr
  • Level: Expert

This role involves designing and optimizing high-performance C++ systems for AI infrastructure, including data pipelines, annotation tooling, and evaluation workflows. The position is ideal for senior C++ engineers who specialize in performance optimization and want to contribute to cutting-edge AI research infrastructure.

  • Data & Machine Learning
  • Platform: Mercor
  • Location: United States
  • Level: Intermediate

This job is an MLOps role on a generative AI team, focused on the performance side of large language model work. The engineer designs and reviews hard tasks on accelerator kernels, profiling, distributed debugging and inference serving, and writes clear technical feedback on solutions produced by others. It suits hands-on systems engineers with production experience in ML infrastructure, model serving or accelerator performance.

Posted October 22, 2026

OpenRecently verified$90-120/hr

Referral link: we may earn a fee. Apply without it

  • Software Engineering
  • Platform: Mercor
  • Location: United States

Mercor is looking for experienced software engineers to evaluate and rate code generated by AI models. You will be asked to write engineering problem statements, examine AI solutions for production robustness, and identify hidden defects. This role is aimed at developers with solid industrial experience and excellent ability to document their technical analysis.

Posted September 15, 2026

OpenRecently verified$70-150/hr

Referral link: we may earn a fee. Apply without it

  • Software Engineering
  • Platform: DataAnnotation

Backend engineers will evaluate how AI models reason about system design, data modeling, and API contracts, identifying failures like race conditions and N+1 queries that only surface under real load. The role involves stress-testing model-generated backend code and writing correct solutions when models fall short, with explicit attention to scaling and consistency trade-offs.

Talent poolRecently verified$40-150/hr
  • Software Engineering
  • Platform: DataAnnotation

Evaluate and stress-test user interfaces built by AI models to identify frontend failures in components, state management, accessibility, and rendering. Write correct implementations when models fall short, providing the training signal through clear technical explanations of what works and what breaks.

Talent poolRecently verified$40-150/hr

Keep exploring

Apply on Mercor (opens Mercor in a new tab)

This is a referral link: we may earn a fee if you are selected and meet the platform's conditions.

Apply without the referral link