Turing Stem Physics

About Turing: Turing is one of the world’s fastest-growing AI companies accelerating the advancement and deployment of powerful AI systems.

Turing helps customers in two ways: Working with the world’s leading AI labs to advance frontier model capabilities in thinking, reasoning, coding, agentic behavior, multimodality, multilinguality, STEM and frontier knowledge; and leveraging that work to build real-world AI systems that solve mission-critical priorities for companies.

Role Overview: This role is part of a project supporting leading LLM companies. The primary objective is to help these foundational LLM companies improve their Large Language Models.

We support companies in enhancing their models by offering high-quality proprietary data. This data can be used as a basis for fine-tuning models or as an evaluation set to benchmark the performance.

In an SFT data generation workflow, you might have to put together a prompt that contains code and questions, then elaborate model responses, and translate the provided CUDA/C++ code into equivalent Python code using PyTorch and NumPy to replicate the algorithm's behavior.

For RLHF data generation, you may need to create a prompt or use one provided by the customer, ask the model questions, and evaluate the outputs generated by different versions of the LLM, comparing it and providing feedback, which is then used in fine-tune processes.

Please note that this role does not involve building or fine-tuning LLMs.

What does day-to-day look like:

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

Requirements:

• 5+ years of overall work experience, with at least 3 years of relevant experience in Python and 2+ years in CUDA/C++.
• Strong hands-on experience with Python, especially in 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 and attention to numerical consistency when porting logic across languages.
• Experience evaluating AI-generated code or participating in LLM tuning is a plus.
• Ability to communicate technical feedback clearly and constructively.
• Fluent in conversational and written English communication skills.

Perks of Freelancing With Turing:

• Work in a fully remote environment.
• Opportunity to work on cutting-edge AI projects with leading LLM companies.

Offer Details:

• Commitments Required: At least 4 hours per day and minimum 20 hours per week with overlap of 4 hours with PST. (We have 3 options of time commitment: 20 hrs/week, 30 hrs/week or 40 hrs/week)
• Engagement type: Contractor assignment (no medical/paid leave)
• Duration of contract: 3 months; [expected start date is next week]
• Location: India, Pakistan, Nigeria, Kenya, Egypt, Ghana, Bangladesh, Turkey, Brazil, Mexico

Evaluation Process (approximately 75 mins): Two rounds of interviews (60 min technical + 15 min cultural & offer discussion)