Job Description

Periodic Labs seeks a Research Scientist to advance reinforcement learning algorithms for frontier AI models in materials and energy discovery. The role involves designing scaling experiments, developing improved RL algorithms, and translating research findings to large-scale production training runs.

Responsibilities

• Design experiments studying RL performance scaling across compute, model size, data, and reward quality
• Develop improved RL algorithms for policy optimization, exploration, and credit assignment in long-horizon tasks
• Build adaptive sampling and curriculum learning methods that dynamically adjust task difficulty and training parameters
• Investigate training stability issues including importance-sampling corrections and policy staleness mitigation
• Optimize compute efficiency through systematic hyperparameter experimentation and infrastructure improvements

Requirements

• 5+ years hands-on experience training language models with reinforcement learning
• Bachelor's degree or equivalent practical experience in relevant field
• Strong scientific rigor and attention to detail in experimental design
• Ability to design small-scale RL experiments that transfer to large-scale settings
• Proficiency working across complex machine learning training stacks

Compensation: $250,000-$350,000 base + equity

Job Tags

Similar Jobs