GPU Kernels AI Task Auditor - Freelance AI Trainer Project
Agency
WorldwideFully remoteContractSeniorInformation Technology
USD 70–100 an hourPosted 8 days ago1w agoApply link checked 14 hours ago
Why we think you can apply
- Hiring model
- Open worldwide · the employer states no country restriction
- Work from
- Anywhere, the UAE included
Posted 8 days ago · apply link checked 14 hours ago
About the role
Responsibilities
- Assess whether GPU kernel tasks are technically accurate, realistic, solvable, reproducible, and supported by reliable tests and evaluation criteria
- Provide clear, actionable feedback on any identified memory allocation bugs, execution bottlenecks, or parallel computing logic errors within the tasks
Requirements
- Demonstrable professional experience and deep technical knowledge in GPU Kernel development (including CUDA/Triton programming, parallel computing, and hardware-level performance optimization)
- Strong analytical and problem-solving skills to rigorously test, troubleshoot, and evaluate complex technical scenarios
About the Company
This project involves sourcing technical specialists to audit tasks used to train and evaluate AI systems to ensure workflows are technically rigorous and accurate.
What we look for
- Deep technical knowledge in GPU Kernel development
- Experience with CUDA/Triton programming
- Parallel computing expertise
- Hardware-level performance optimization skills
- Strong analytical and problem-solving skills
CUDATritonGPU KernelsParallel Computing
SponsorWake up to a shortlist, not a search results page.
NeverApply scores every new listing against your CV, salary floor and visa. A handful of real matches by morning.
Get your daily matchesSimilar jobsbased on title, category and scope
Agency
SponsorWake up to a shortlist, not a search results page.
NeverApply scores every new listing against your CV, salary floor and visa. A handful of real matches by morning.
Get your daily matchesGPU Kernels AI Task Auditor - Freelance AI Trainer ProjectAgency · free account, 30 seconds