Research Engineer, AGI Safety and Alignment

Salary: £160,000 - 200,000 per year

Requirements:

• We require a bachelors degree in Computer Science, a related Software Engineering field, or equivalent practical experience.
• We require 3 years of experience in software development, ML engineering, or ML research.
• We require experience working with research teams.
• Preferred: experience conducting or contributing to applied research to improve the safety and alignment of frontier AI systems.
• Preferred: experience training large models, such as supervised fine-tuning or RLHF.

Responsibilities:

• We research new alignment methods, study alignment failures, and apply AGI-scalable alignment techniques to frontier models.
• We develop adversarially robust AGI control systems and implement them in production.
• We research interpretability techniques to understand what AI systems are thinking.
• We work with product teams to ensure our research is correctly adopted.
• We research novel techniques to reduce existential and catastrophic risk from AGI and ASI.
• We advise executive leadership on safety and risks posed by AI systems.
• We help find and fix sources of misalignment, explore alignment techniques with better generalization, and build tools that accelerate safety research.

Technologies:

• AI
• Fine-tuning
• REST

More:

We are the Artificial General Intelligence (AGI) Safety and Alignment Team (ASAT) at Google DeepMind, and our mission is to reduce existential and catastrophic risk from AGI and eventually ASI. We are a pioneering AI lab focused on advancing AI to solve complex global challenges while ensuring safety and ethics remain our highest priorities. Our team works across Google DeepMind and Google to apply novel safety techniques, and we advise executive leadership on AI safety. We are prioritising hires in deep alignment, alignment stress testing, language model interpretability, agent control, and amplified oversight, with opportunities available for both Research Scientists and Software Engineers depending on background.

last updated 34 week of 2026