Research Scientist/Engineer (Evaluations)

Core responsibilities:

• Run pre-deployment evaluation campaigns on the most capable AI systems in the world, partnering with multiple labs to access a breadth of models
• Deep dive into AI cognition by scanning through thousands of model transcripts to surface behavioral patterns
• Build new evaluations for frontier risks, from designing novel test environments to scaling them across hundreds of distinct scenarios
• Automate and improve the evaluation pipeline, rethinking and reshaping it as new possibilities emerge

This role helps assess the risks posed by scheming AIs, with evaluations directly informing deployment decisions for the most capable AI systems in the world, contributing to AI safety and responsible development.

Required skills and qualifications:

• Strong software engineering skills with experience shipping and maintaining production Python code
• Process optimization skills with a focus on improving workflows and reducing friction
• Data analysis and pattern recognition abilities to extract signal from large, messy datasets
• Excellent writing and communication skills to convey findings to both technical and non-technical audiences
• AI power-user with experience using different models and experimenting with new AI workflows

No specific benefits, compensation, or salary information mentioned in the job posting.

Apollo Research is primarily concerned with risks from Loss of Control in AI, particularly deceptive alignment/scheming, and works on detection, science, and mitigation of these risks through evaluations and collaboration with frontier AI companies.