Research Engineer - Posttraining
Job Description
- Creating and scaling RL environments for LLMs
- Creating high-quality evals for frontier models
- Working closely with domain experts to define evaluation criteria, tools, and environments for agents
- Carefully crafting training datasets and reward functions, with LLMs and/or human trainers
- Training frontier LLMs with RL
About Periodic Labs
We are an AI + physical sciences lab building state of the art models to make novel scientific discoveries. We are well funded and growing rapidly. Team members are owners who identify and solve problems without boundaries or bureaucracy. We eagerly learn new tools and new science to push forward our mission.
About the role
In this role, you will post-train frontier models to autonomously run various parts of the scientific discovery pipeline. Models you train will generate hypotheses, design experiments that run in an actual lab, operate sophisticated scientific equipment, and more. You will work with the world’s leading experts in the physical sciences in order to create high-quality evaluation and training tasks, scale up RL environments, design creative reward functions, and run large-scale RL runs, all in service of automating scientific discovery.
You might thrive in this role if you have experience:
- Creating and scaling RL environments for LLMs
- Creating high-quality evals for frontier models
- Working closely with domain experts to define evaluation criteria, tools, and environments for agents
- Carefully crafting training datasets and reward functions, with LLMs and/or human trainers
- Training frontier LLMs with RL
