Research Scientist

Description

About the Role

Join a post-training research team focused on measuring how training data affects model behavior. You will run controlled SFT and reinforcement learning experiments, then turn the results into practical recommendations for improving data quality and model performance.

Responsibilities

• Design, run, and analyze controlled SFT and RL experiments to assess dataset impacts on model capabilities, generalization, and alignment.
• Help build evaluation infrastructure and benchmarks that make results reproducible across models, capabilities, and domains.
• Develop public evaluations and data types that support frontier model research.
• Publish research, technical reports, and blog posts that communicate findings clearly.
• Collaborate with internal partners and research labs to improve data quality based on experimental results.

Requirements

• Zero to four years of experience building or delivering work in LLM training, post-training, or model evaluation.
• Experience at a frontier AI lab and strong familiarity with LLM training and evaluation methods.
• Ability to independently design, implement, and analyze controlled experiments, including when results are messy or inconclusive.
• Software engineering skills to build or modify experimental infrastructure, such as evaluation harnesses, training pipelines, or benchmarks.
• Experience working across domains such as finance, software engineering, and policy.
• A practical, builder-oriented approach; meaningful research or end-to-end post-training experiments are valuable.

Conditions

Base salary: $150,000 to $250,000 annually. Visa sponsorship is available.

On-site in San Francisco, California, United States.