LLM Trainer – CUDA/C++ to Python at Turing

LLM Trainer – CUDA/C++ to Python Migration | Contractor | Remote (Select Countries)

Join Turing's AI data generation team and help improve the world's most advanced Large Language Models by translating CUDA/C++ code into Python using PyTorch and NumPy. This is a fully remote contractor role ideal for experienced GPU/systems programmers looking to contribute to cutting-edge AI development.

About Turing

Turing is one of the world's fastest-growing AI companies, partnering with leading AI labs to advance frontier model capabilities in reasoning, coding, STEM, and more. We provide high-quality proprietary data to help foundational LLM companies fine-tune and benchmark their models.

Role Overview

This role supports LLM improvement through SFT (Supervised Fine-Tuning) and RLHF (Reinforcement Learning from Human Feedback) data generation workflows. You will translate CUDA/C++ code into equivalent Python implementations and evaluate AI-generated code translations — helping models learn from expert-level technical feedback.

Note: This role does not involve building or fine-tuning LLMs directly.

Day-to-Day Responsibilities

• Translate CUDA/C++ code into Python using PyTorch and NumPy, ensuring logical and performance parity.
• Analyze CUDA kernels and GPU-accelerated code for structure, efficiency, and function prior to translation.
• Evaluate LLM-generated CUDA/C++ to Python translations, providing detailed technical feedback and corrections.
• Collaborate with prompt engineers and researchers to develop test prompts reflecting real-world CUDA/PyTorch tasks.
• Participate in RLHF workflows — rank LLM responses and justify ranking decisions clearly.
• Debug and review translated Python code for correctness, readability, and adherence to industry standards.
• Maintain technical documentation to support reproducibility and code clarity.
• Propose improvements to prompt structure or conversion approaches based on observed LLM failure patterns.

Requirements

• 5+ years of overall work experience, including at least 3 years in Python and 2+ years in CUDA/C++.
• Strong hands-on experience with scientific computing using PyTorch and NumPy.
• Solid understanding of CUDA programming concepts and C++ fundamentals.
• Demonstrated ability to analyze CUDA kernels and accurately reproduce them in Python.
• Familiarity with GPU computation, parallelism, and performance-aware coding practices.
• Strong debugging skills with attention to numerical consistency when porting logic across languages.
• Experience evaluating AI-generated code or participating in LLM tuning workflows is a plus.
• Fluent in written and conversational English.

Engagement Details

• Type: Contractor assignment (no medical/paid leave benefits)
• Duration: 3 months (with expected start next week)
• Commitment: Minimum 20 hrs/week (options: 20, 30, or 40 hrs/week); at least 4 hours/day with 4-hour overlap with PST
• Location: India, Pakistan, Nigeria, Kenya, Egypt, Ghana, Bangladesh, Turkey, Brazil, Mexico

Perks

• Fully remote work environment
• Opportunity to contribute to cutting-edge AI projects with leading LLM companies

Evaluation Process (~75 minutes)

Two rounds: a 60-minute technical interview followed by a 15-minute cultural fit and offer discussion.