CUDA C++ to Python AI Code Evaluation Expert

About OpenTrain

OpenTrain is the #1 platform for finding and building careers in AI training and data labeling. We help contributors discover specialized projects, build a profile that reflects their expertise, and apply to opportunities in a fast-growing industry.

For this role, OpenTrain AI is hiring a remote contractor to support high-quality coding data development and evaluation for large language models.

• Remote contractor opportunity
• Part-time engagement of 20+ hours per week
• Apply and build your AI training profile through OpenTrain

About AI Code Evaluation

AI training is the human side of building artificial intelligence. Coding experts help models improve by creating programming examples, reviewing generated solutions, comparing responses, and explaining what makes code correct, efficient, and reliable.

This role focuses on producing and evaluating coding data rather than building or fine-tuning language models. Your technical judgment will help assess how well AI-generated code handles real CUDA, C++, Python, and scientific-computing tasks.

• Contribute to cutting-edge AI development
• Use specialized programming expertise in flexible remote work
• Evaluate model responses using clear technical standards

The Role

As a CUDA/C++ to Python AI Code Evaluation Expert, you will translate CUDA and C++ implementations into equivalent Python code using PyTorch and NumPy. You will assess whether translations preserve the original behavior, numerical consistency, and relevant performance characteristics.

The work also includes reviewing generated code, creating practical programming prompts, ranking alternative model responses, and providing precise technical feedback. The listing identifies this opportunity as entry level, while the role requires substantial professional experience in Python, CUDA, and C++.

• Work remotely as a contractor
• Commit 20+ hours per week
• Support coding data for large language model evaluation
• Use English for written and conversational communication

What You'll Do

You will combine implementation, debugging, evaluation, and technical communication. Your reviews should be accurate, reproducible, and clear enough to explain both the problem and the appropriate correction.

• Translate CUDA and C++ code into equivalent Python using PyTorch and NumPy
• Preserve logical and numerical consistency across implementations
• Analyze CUDA kernels and GPU-accelerated code for structure, efficiency, behavior, and parallelism
• Review generated code translations and identify correctness and performance issues
• Recommend improvements to translated implementations
• Create or refine prompts for practical CUDA and PyTorch programming tasks
• Compare and rank model responses using clear technical criteria
• Explain evaluation decisions and provide precise technical feedback
• Debug translated Python code for correctness, readability, reproducibility, and quality
• Maintain technical documentation and suggest improvements based on recurring coding errors

Required Skills and Experience

You should bring at least five years of overall professional experience, including at least three years of Python experience and at least two years working with CUDA and C++. Strong hands-on scientific-computing experience with PyTorch and NumPy is required.

You must be able to analyze CUDA kernels, reproduce their behavior accurately in Python, debug numerical issues, and communicate technical feedback clearly. Experience evaluating AI-generated code or contributing to language-model tuning workflows is helpful but not required.

• At least five years of overall professional experience
• At least three years of relevant Python experience
• Hands-on scientific-computing experience with PyTorch and NumPy
• At least two years of experience with CUDA and C++
• Knowledge of GPU computation, parallelism, and performance-aware coding
• Ability to preserve behavior and numerical consistency when translating code
• Strong debugging and generated-code evaluation skills
• Ability to explain technical decisions and provide clear corrections
• Fluent written and conversational English

Who Should Apply

This opportunity is suited to experienced Python and GPU-programming practitioners who enjoy examining implementations closely and explaining technical decisions. It may be a strong fit if you have worked with scientific computing, CUDA kernels, parallel code, or performance-sensitive systems and want to apply that expertise to AI training.

Experience with AI-generated code evaluation or language-model tuning workflows can help you contribute, but the core requirement is demonstrated technical strength in CUDA, C++, Python, PyTorch, and NumPy.

• CUDA and C++ developers with strong Python skills
• Scientific-computing practitioners familiar with PyTorch and NumPy
• Engineers who can reason about GPU parallelism and numerical behavior
• Technical reviewers who communicate corrections clearly

Eligible Locations and How to Apply

This remote opportunity is open to contractors located in India, Pakistan, Nigeria, Kenya, Egypt, Ghana, Bangladesh, Türkiye, Brazil, or Mexico. OpenTrain provides a place to present your relevant experience, discover AI training work, and apply in minutes.

Creating an OpenTrain account is free. Build a profile that showcases your technical background and use it to pursue opportunities that match your skills.

• Eligible countries: India, Pakistan, Nigeria, Kenya, Egypt, Ghana, Bangladesh, Türkiye, Brazil, and Mexico
• Language: English
• Schedule: 20+ hours per week
• Engagement: part-time contractor
• Remote work with OpenTrain AI