Research Scientist, AGI Safety and Alignment, DeepMind

Overview

In this role you will advance AGI safety and alignment research, translating methods into production-ready systems. You will collaborate with cross-functional teams to ensure research is adopted in products, and explore interpretability to understand AI decisions. You will shape defense-in-depth controls for AGI deployments, advise leadership on risk, and contribute to scalable alignment efforts for frontier models.

Responsibilities

• Research new alignment methods for frontier models and study alignment failures
• Develop AGI control systems and implement them in production
• Investigate interpretability techniques to understand AI system thinking
• Collaborate with product teams to ensure research is adopted in products
• Advise executive leadership on safety risks using frontier safety framework
• Contribute to alignment stress testing, evaluative research, and tool development
• Explore training techniques like debate for aligning superhuman AI
• Support safety research with tools for model forensics and eval awareness
• Participate in cross-functional efforts to reduce existential and catastrophic AI risk

Key requirements

• Bachelor’s degree in Computer Science, related Software Engineering field, or equivalent practical experience
• 3 years of experience in software development, ML engineering, or ML research
• Experience working with research teams

• collaboration with cross-functional teams
• clear technical communication
• problem solving under high risk contexts
• ML research and engineering experience
• training large models (supervised finetuning, RLHF)
• interpretability research

…