Research Scientist, Frontier Computation, DeepMind

Overview

As a Research Scientist on the Frontier Computation team, you will lead long-term moonshot research and guide direct product deployments. You probe model internals, pursue mechanistic interpretability, and study information propagation with high initiative, designing rigorous empirical experiments. You own AI safety and security engagements, developing prompt defenses and alignment strategies for frontier systems. You translate complex findings into structured results and collaborate closely with internal and external partners to advance foundational, cross-cutting research.

Responsibilities

• Lead moonshot research programs and guide deployments on DeepMind AI offerings
• Probe LLM internals, mechanistic interpretability, and information propagation
• Design and execute rigorous empirical experiments with clear baselines and ablations
• Own AI safety and security engagements, developing prompt defenses and alignment strategies
• Translate findings into structured results and communicate with collaborators
• Collaborate with internal and external partners to advance research directions
• Contribute foundational work across multiple frontier research projects
• Maintain a continuous learning mindset and broaden contributions across GDM domains

Key requirements

• PhD in Computer Science or related field, or equivalent practical experience
• 2 years of experience with Large Language Model internals or generalization

• collaborative working style
• independence and initiative
• strong experimental judgment
• LLM internals
• mechanistic interpretability
• information propagation

…