Research Scientist, Frontier Risk Evaluations
The Research Scientist, Frontier Risk Evaluations, will join Scale Labs, Scale AI's research division, focusing on policy research to bridge AI research and global policymakers. This role involves designing evaluation measures and datasets to assess risks posed by advanced AI systems, collaborating with government agencies and other labs to scope and design evaluations, and publishing methodologies and technical reports for policymakers.
Key responsibilities include building harnesses to test AI models for dangerous capabilities, such as security vulnerabilities and high-risk activities, and working with cross-functional teams to develop and implement evaluation strategies. The role also involves publishing findings and contributing to the development of AI safety policies.
Candidates should have at least three years of experience addressing sophisticated machine learning problems, with a track record of published research in machine learning, particularly in generative AI. Practical experience in conducting technical research collaboratively, building and instrumenting machine learning pipelines, and writing evaluation harnesses is essential. Strong written and verbal communication skills are required to operate in a cross-functional team.
The base salary range for this full-time position in San Francisco and New York is $197,400 to $246,750 USD. Compensation packages include base salary, equity, and benefits such as comprehensive health, dental, and vision coverage, retirement benefits, a learning and development stipend, and generous paid time off.
Scale AI is a leading data and evaluation partner for frontier AI companies, specializing in understanding and safeguarding AI models and systems. The company collaborates broadly across industry, the public sector, and academia, regularly publishing findings to inform AI risk evaluations. Joining Scale AI offers an opportunity to contribute to the advancement of AI safety and policy research in a dynamic and growing organization.