Remote GPU Kernel Auditor for AI Task Evaluation

Jobgether; Poland; On-site; PLN 358,000 - 511,000; Full time

Benefits offered by this job

Competitive pay

Fully remote (India)

Contractor flexibility

High-impact technical work

Specialized focus

Job summary

Jobgether, on behalf of a partner company, is seeking a GPU Kernels AI Task Auditor - Freelance AI Trainer based in Poland. This remote role invites experienced GPU kernel specialists to apply technical expertise to the development and evaluation of AI systems.

You will audit GPU kernel tasks used to train and assess AI models, test memory allocation, synchronization, and parallel computing logic, and provide clear actionable feedback. The engagement is fully remote and contract-based.

Qualifications

• Demonstrable experience in GPU kernel development and optimization.
• Strong knowledge of CUDA or Triton programming and parallel computing.
• Ability to identify memory management and synchronization issues.
• Clear written communication of technical findings.

Responsibilities

• Evaluate GPU kernel tasks for technical accuracy, realism, solvability, and reproducibility.
• Test and troubleshoot complex GPU programming scenarios.
• Audit tasks for memory allocation, bottlenecks, and parallel execution issues.
• Review CUDA, Triton, and other GPU kernel implementations for correctness and performance.
• Provide actionable feedback and improvement recommendations.
• Work independently within an assigned specialization area.

Skills

GPU kernel development

CUDA/Triton programming

Parallel computing

Hardware-level GPU optimization

Analytical problem solving

Technical feedback writing

Tools

CUDA

Triton

Job description

Jobgether, on behalf of a partner company, is seeking a GPU Kernels AI Task Auditor - Freelance AI Trainer based in Poland. This remote role invites experienced GPU kernel specialists to apply technical expertise to the development and evaluation of AI systems.

You will audit GPU kernel tasks used to train and assess AI models, test memory allocation, synchronization, and parallel computing logic, and provide clear actionable feedback. The engagement is fully remote and contract-based.