Research Scientist, AI Safety and Alignment
$174k - $252k • Seattle, WA, USA • Full-time
Posted 8d ago
Job Location
Seattle, WA, USA
Tech Stack
Remote Work Policy
On-site
Employment Type
Full-time
Categories
AI Research Engineer
About the job
Google DeepMind's Deep Understanding team is seeking a Research Scientist focused on AI Safety and Alignment. This role involves empirical science and practical engineering to address concrete safety wins and solve real vulnerabilities in live systems, particularly for frontier models. The team prioritizes a sustainable, high-trust research environment where independent thinking is encouraged, and success is measured by the safety and reliability of AI models. As a Research Scientist, you will contribute to the wider research community by sharing and publishing findings, collaborating with universities and technical institutes globally, and helping to develop long-term research strategies to expand Google's research impact.
Responsibilities
- Define data structure, framework, design, and evaluation metrics for research solution development and implementation, identifying timelines and resources.
- Identify and address problems or gaps in existing technology by engaging stakeholders and leaders.
- Author research papers to share and generate impact of research results within the function and the broader research community.
- Identify new research areas through collaboration with internal and external partners, contributing to long-term research strategy.
- Share research trends and best practices within the community to help grow the research business.
Requirements
- PhD in Computer Science, Machine Learning, Artificial Intelligence, Mathematics, Physics, or a related STEM discipline, or equivalent practical experience.
- 1 year of experience in AI safety, alignment, or mechanistic interpretability research.
- Experience evaluating, steering, red teaming, or implementing empirical safeguards on LLMs or frontier systems.
- Proficiency in Python or C++, deep learning frameworks (e.g., JAX, PyTorch), and agentic workflows or tools.
- First-author publications at major machine learning conferences (e.g., NeurIPS, ICML, ICLR, COLM).
Benefits
- 15% bonus target
- Equity
- Benefits package
