CUDA GPU Kernel Optimization Engineer

About OpenTrain

OpenTrain is the #1 platform for finding and building careers in AI training and data labeling. We help people discover specialized projects, build a professional profile, and grow their experience in the rapidly expanding AI industry.

About AI Training Work

AI training is the human side of building artificial intelligence. In addition to labeling data and evaluating model outputs, this work can include writing and reviewing code, testing technical approaches, and applying specialized expertise to improve the systems behind modern AI.

• Remote work that can fit around other commitments
• Opportunities to apply specialized technical expertise to cutting-edge AI development
• A chance to build a portfolio of practical AI training and evaluation experience

The Role

OpenTrain is hiring a CUDA GPU Kernel Optimization Engineer to support AI training work centered on GPU performance and real-world technical expertise. You will optimize GPU kernels, improve C++ and CUDA code, develop shader workflows, and evaluate GPU-based approaches using measurable performance analysis.

This is a remote, part-time contractor role requiring 20+ hours per week. Compensation is $60–$100 per hour. Prior AI experience is not required; the focus is on GPU programming and performance-engineering expertise.

• Employment type: Contractor and part-time
• Work arrangement: Remote, worldwide
• Time requirement: 20+ hours per week
• Compensation: $60–$100 per hour
• Primary language: English

What You’ll Do

You will investigate GPU performance challenges, implement optimization strategies, and communicate your findings through clear technical documentation. The role combines hands-on programming with analytical evaluation across modern GPU hardware.

• Analyze, profile, and optimize GPU kernels to increase computational throughput on modern hardware.
• Identify kernel bottlenecks and recommend targeted optimization strategies.
• Refactor C++ and CUDA code for maintainability, efficiency, and adaptability across GPU architectures.
• Implement GLSL and WebGPU shader logic and graphics or compute workflows.
• Document optimization steps, findings, and performance improvements in clear technical reports.
• Contribute to design discussions and evaluate GPU performance metrics and approaches.
• Track developments in GPU programming and share relevant technical insights.

Required Skills

This role is designed for candidates with demonstrated GPU programming and high-performance development experience. Strong analytical thinking and communication are important for explaining performance tradeoffs and documenting technical decisions.

• Demonstrated CUDA programming expertise and experience tuning GPU kernel performance.
• Advanced C++ development skills in high-performance computing environments.
• Hands-on GLSL and WebGPU experience for graphics or compute shader development.
• Proficiency with GPU profiling tools such as Nsight, Visual Profiler, or comparable tools.
• Strong analytical ability to reason about kernel performance across hardware generations.
• Clear written and verbal technical communication.
• Ability to produce clear technical documentation, reporting, and analysis.

Helpful Background

Experience collaborating in remote, cross-disciplinary project settings is helpful. The listed experience level is entry level, but the role requires the specific technical capabilities described above, including CUDA performance tuning, advanced C++, shader development, and GPU profiling.

Build Your AI Training Career

OpenTrain gives contributors a place to develop a credible AI training profile and find projects that match their expertise. By contributing specialized technical work, you can build experience in how advanced AI systems are developed and evaluated.

• Create an OpenTrain account for free.
• Showcase your GPU programming and performance-engineering expertise.
• Discover relevant AI training and technical evaluation opportunities.
• Apply in minutes and continue building your professional portfolio.