Research Scientist Scaling RL
AI research and deployment company building AI scientists and autonomous laboratories for physical-science discovery.
About Periodic Labs
Periodic Labs develops specialized AI models and high-throughput autonomous labs that run and analyze physical experiments, initially for materials discovery including superconductors, magnets, and semiconductor applications.
Skills
Candidate Availability
Required and preferred rules are kept separate and reflect the wording in the original posting.
About the Role
You will study reinforcement-learning scaling, design controlled experiments, develop algorithms for long-horizon tasks, and transfer research ideas to large-scale training runs. You will build adaptive sampling and curriculum methods, investigate training bias and stability, and improve training and inference efficiency through hyperparameter experiments.
Requirements
- Hands-on experience training LLMs with reinforcement learning
- Strong attention to detail and a rigorous scientific approach
- Ability to develop small-scale reinforcement-learning setups that transfer to large-scale training runs
- Ability to implement, debug, and test research ideas across a complex training stack
- 5+ years of experience
- Bachelor’s degree or similar experience
Responsibilities
- Design experiments to understand how reinforcement-learning performance scales with compute, model size, data, and reward quality
- Develop reinforcement-learning algorithms for policy optimization, advantage estimation, exploration, and credit assignment
- Build adaptive sampling and curriculum methods
- Study bias and stability during reinforcement-learning training
- Improve compute efficiency across training and inference through hyperparameter experiments
Benefits
- Equity
- Visa sponsorship