GPU Kernels AI Task Auditor - Freelance AI Trainer Project in United Arab Emirates at Jobgether

This position is listed on behalf of a partner company, who manages all applications and next steps. Our partner is looking for a GPU Kernels AI Task Auditor - Freelance AI Trainer Project based in United Arab Emirates.

This freelance opportunity is designed for experienced GPU kernel specialists who want to apply their technical expertise to the development and evaluation of AI systems.
You will audit GPU kernel tasks used to train and assess AI models, ensuring they meet high standards of technical accuracy and practical relevance.
The role focuses on evaluating whether complex tasks are realistic, solvable, reproducible, and supported by robust testing and evaluation criteria.
You’ll investigate potential memory allocation issues, execution bottlenecks, and parallel computing logic errors.
Your feedback will help improve the quality, reliability, and technical rigor of AI training workflows.
The position offers fully remote work and the opportunity to contribute to cutting-edge AI development through specialized GPU expertise.

• Evaluate GPU kernel tasks to determine whether they are technically accurate, realistic, solvable, reproducible, and supported by reliable tests and evaluation criteria.
• Rigorously test and troubleshoot complex GPU programming scenarios to validate task behavior and identify potential technical weaknesses.
• Audit tasks for memory allocation bugs, execution bottlenecks, synchronization issues, and parallel computing logic errors.
• Review CUDA, Triton, and other GPU kernel implementations for correctness, performance considerations, and practical applicability.
• Provide clear, specific, and actionable technical feedback on identified issues and recommend improvements where appropriate.
• Apply deep knowledge of GPU computing and hardware-level optimization to assess the quality and relevance of AI training and evaluation tasks.
• Work independently within an assigned area of specialization while maintaining a high standard of technical accuracy and consistency.

• Demonstrable professional experience and deep technical knowledge in GPU kernel development and optimization.
• Strong expertise in areas such as CUDA or Triton programming, parallel computing, and hardware-level GPU performance optimization.
• Strong analytical and problem-solving skills, with the ability to rigorously evaluate and troubleshoot complex technical scenarios.
• Ability to identify subtle correctness, performance, memory management, and parallel execution issues.
• Strong understanding of practical GPU development workflows and the ability to assess whether technical tasks can be reliably reproduced.
• Clear written communication skills, with the ability to turn technical findings into concise and actionable feedback.
• Candidates will be evaluated for one specialty based on their experience; expertise across every GPU-related domain is not required.
• Ability to work effectively and independently in a fully remote freelance environment.
• Access to a secure computer and high-speed internet connection is required.

• Competitive freelance compensation: $70–$100 USD per hour, with the final rate determined based on experience, expertise, and geographic location.
• Fully remote: Work remotely from India as part of a globally distributed technical project.
• Flexible freelance engagement: Contribute your specialized expertise through a remote contract arrangement.
• High-impact technical work: Help improve the accuracy, reliability, and technical sophistication of AI training and evaluation systems.
• Specialized focus: Work within the GPU kernel specialty that best matches your professional background.
• Contractor flexibility: Designed for experienced specialists seeking project-based technical work.
• No employer-sponsored benefits: Health insurance, paid time off, and other company-sponsored employee benefits do not apply to freelance contractors.
• Equipment responsibility: Contractors provide their own secure computer and high-speed internet connection.