Reinforcement Learning Engineer

At Code Metal AI, you’ll be part of a world class team with talent from MIT, OpenAI and other top companies, focused on pioneering work in large language models (LLMs) and code generation. Our projects directly involve leading chip manufacturers, applying advanced AI to solve meaningful, practical challenges with real world impact.

This role bridges two critical areas:

Production

• Build and maintain robust distributed training systems using PyTorch (2+ years experience required).
• Design and implement scalable data curation and quality assurance pipelines to ensure top-tier training datasets.
• Develop orchestration tools that manage complex workflows across large-scale AI model training and evaluation.

Research

• Drive innovation by developing evaluation frameworks and reinforcement learning solutions, including recent advancements in Reinforcement Learning with Human Feedback (RLHF).
• Engage with frontier research through open-source projects and potential publications, applying RLHF to Large Language Models (LLMs), ideally focusing on code generation tasks.

• 2+ years experience in distributed training, preferably with PyTorch.
• Strong background in reinforcement learning, with recent RLHF experience highly preferred.
• Proven ability to build data curation and quality assurance pipelines.
• Experience with evaluation framework development.
• Ideally, experience across both data pipeline and orchestration sides.
• Eligible for TS/SCI clearance.

Nice to have:

• Contributions to open-source AI or ML projects.
• Published work or demonstrable research experience in related fields.
• Hands-on experience applying RLHF to LLMs, especially for code generation.
• Experience with large scale synthetic data generation.

• Health care plan with 100% premium coverage, including medical, dental, and vision.
• 401k with 5% matching.
• Paid Time Off (Uncapped Vacation, plus Sick & Public Holidays).
• Flexible hybrid work arrangement.
• Relocation assistance for qualifying employees.