AI Safety & Alignment Research Scientist

Google LLC; Seattle (WA); On-site; USD 174,000 - 252,000; Full time

Benefits offered by this job

Health insurance

Dental insurance

Vision insurance

Life insurance

Disability insurance

401(k) match

Paid time off

Sick time

Maternity leave

Baby bonding leave

Holidays

Job summary

Google DeepMind in Seattle, WA seeks a Research Scientist for AI safety and alignment to advance scalable, production-grade safety measures for frontier models. You will design experiments, study model behaviors, and publish findings with collaborations across disciplines.

Lead efforts on interpretability, post-training alignment, and empirical safeguards, contributing to real-world deployments while maintaining a high-trust research environment and strong emphasis on safety and ethics.

Qualifications

• PhD in CS/ML/AI/Math/Physics or related STEM field or equivalent practical experience.
• 1 year of experience initiating, scoping, and executing research in AI safety, alignment, or mechanistic interpretability.
• Experience evaluating, steering, red-teaming, or implementing empirical safeguards on LLMs or frontier systems.
• Experience in Python or C++, deep learning frameworks (e.g., JAX, PyTorch), and agentic workflows or tools.

Responsibilities

• Drive project work by defining data structures, framework, design, and evaluation metrics for research solution development.
• Identify problems in existing technology and engage stakeholders to address them.
• Author research papers to share and generate impact across function and community.
• Identify new research areas by collaborating with external/internal partners and plan long-term strategy.
• Grow research business by sharing trends and best practices within the community.

Skills

Python/C++

Deep learning frameworks

Agentic workflows

First-author papers

Education

PhD in CS/ML/AI/Math/Phys or related

Tools

PyTorch

JAX

Job description

Google DeepMind in Seattle, WA seeks a Research Scientist for AI safety and alignment to advance scalable, production-grade safety measures for frontier models. You will design experiments, study model behaviors, and publish findings with collaborations across disciplines.

Lead efforts on interpretability, post-training alignment, and empirical safeguards, contributing to real-world deployments while maintaining a high-trust research environment and strong emphasis on safety and ethics.