Research Scientist, Safety Post-Training — AI Risk & Policy

Scale AI; San Francisco, New York (CA, NY); On-site; USD 216,000 - 270,000; Full time

Benefits offered by this job

Health

Dental coverage

Vision coverage

Retirement benefits

Learning & development stipend

Generous PTO

Commuter stipend

Job summary

Scale Labs seeks a Research Scientist focused on Safety Post Training to develop post-training methods and interpretability techniques for safer frontier AI systems. You will work on pipelines and evaluations to reveal unsafe behaviors and guide mitigations, collaborating with policymakers, engineers, and researchers.

Ideal candidates have three+ years addressing complex ML problems, with track records in post-training research and strong communication across teams.

Qualifications

• At least three years of experience addressing sophisticated ML problems in research or product development.
• Experience with post-training methods and interpretability evaluations to improve model safety and alignment.
• Experience with RLHF, DPO, GRPO or similar post-training techniques.
• Strong written and verbal communication skills for cross-functional collaboration.

Responsibilities

• Design and run post-training pipelines to study how training choices affect model safety, robustness, and alignment properties.
• Develop interpretability-informed evaluations to reveal unsafe or undesirable model behaviors and guide mitigations.
• Collaborate with policymakers, engineers, and researchers to translate findings into safety standards, evaluation benchmarks, and best practices.

Job description

Scale Labs seeks a Research Scientist focused on Safety Post Training to develop post-training methods and interpretability techniques for safer frontier AI systems. You will work on pipelines and evaluations to reveal unsafe behaviors and guide mitigations, collaborating with policymakers, engineers, and researchers.

Ideal candidates have three+ years addressing complex ML problems, with track records in post-training research and strong communication across teams.