As models become more agentic, executing long-horizon tasks, using tools, writing and running code, operating across multi-step workflows, the challenge of making them safe fundamentally changes. Surface-level safety methods (output filtering, refusal tuning, policy guardrails) were designed for single-turn interactions. They are not enough for agents that plan, act, and adapt over extended horizons.
We are looking for a Senior Staff Research Scientist to rethink safety post-training for this new reality. You will bring frontier post-training expertise, to develop training methods that make Gemini models deeply safe and aligned, especially in agentic settings. This role sits in Gemini Safety and partners closely with the Artificial General Intelligence (AGI) Safety team and the Gemini post-training organization.
Artificial intelligence will be one of humanity’s most transformative inventions. At Google DeepMind, we are a pioneering AI lab with exceptional interdisciplinary teams focused on advancing AI development to solve complex global challenges and accelerate high-quality product innovation for billions of users. We use our technologies for widespread public benefit and scientific discovery, ensuring safety and ethics are always our highest priority.
We are pushing the boundaries across multiple domains. Our global teams offer learning opportunities and varied career pathways
for those driven to achieve exceptional results through collective effort.
Individual pay is determined by factors including job-related skills, experience, and relevant education or training.
US: $262000 - $365000 (USD) + 25% bonus target + equity + benefits
Learn more about
benefits at Google.
Responsibilities
- Rethink how safety is trained into models, especially for agentic, long-horizon behavior.
- Design and ship post-training recipes (Reinforcement Learning (RL), Supervised Fine-Tuning (SFT), and beyond) that install safety and alignment properties into Gemini models. You own the path from research to production.
- Build the metrics and evaluations that tell us whether training is actually making models safer in deployment, not just on benchmarks.
- Work directly with the post-training pipeline and infrastructure. Partner with the AGI Safety team to bring alignment research into practical training. Translate between research and production.
- Shape the road map for where safety post-training goes next. Build and grow the team to execute on it.
Minimum qualifications:
- PhD in Computer Science, a related field, or equivalent practical experience.
- 6 years of experience in Machine Learning Algorithms and Language Modeling.
- One or more scientific publications in the ML/AI conferences or journals (e.g., NeurIPS, ICML, ICLR, CVPR).
Preferred qualifications:
- 5 years of experience in safety/alignment, including RLHF, reward modeling, and out-of-model safety systems. Proven track record of mitigating model risks at scale.
- 5 years of documented experience driving research concepts from initial hypothesis through to product realization.
- Experience designing and deploying AI agents and safety-critical, high-availability systems.
- Expertise in designing/executing comprehensive model evaluation frameworks to identify, quantify, and close critical safety gaps.
- Deep technical experience across the full LLM life-cycle, including pre-training, inference optimization, and fine-tuning.