Research Scientist, Evaluations, Security and Privacy, Deepmind

Google Google · Big Tech · Mountain View, CA +2

Research Scientist focused on security and privacy for AI models and agentic products, specifically Gemini. The role involves designing and evaluating novel defense mechanisms against adversarial attacks and prompt injections, translating research into practical solutions for training and inference pipelines, and collaborating with core modeling and engineering teams. The position requires a PhD and experience in ML research, benchmarking, and security, with a focus on next-generation security techniques for autonomous AI systems.

What you'd actually do

  1. Drive research to safeguard Gemini’s flagship foundational models and agentic products against emerging vulnerabilities at a massive scale.
  2. Design, prototype, and evaluate novel defense mechanisms to protect models and agents from adversarial attacks, prompt injections, and contextual security threats.
  3. Translate theoretical research breakthroughs into practical, real-world security solutions for both training and inference pipelines.
  4. Work closely with core modeling, engineering, and Trust and Safety teams to seamlessly integrate security innovations into Gemini's infrastructure.
  5. Stay ahead of the threat landscape by inventing next-generation security techniques specifically designed for autonomous and agentic AI systems.

Skills

Required

  • PhD degree in Computer Science, a related field, or equivalent practical experience
  • 4 years of experience with research agendas across multiple teams or projects
  • 3 years of experience designing and implementing benchmarking frameworks for machine learning models
  • 2 years of experience in security and privacy
  • One or more scientific publication submissions for conferences, journals, or public repositories (such as CVPR, ICCV, NeurIPS, ICML, ICLR, etc.)

Nice to have

  • 3 years of experience in software development or engineering
  • 2 years of experience coding in C++ and Python
  • Passion for AI technology and all of its possibilities

What the JD emphasized

  • security and privacy
  • agentic products
  • adversarial attacks
  • prompt injections
  • contextual security threats
  • autonomous and agentic AI systems
  • security innovations
  • emerging vulnerabilities
  • novel defense mechanisms
  • benchmarking frameworks for machine learning models
  • security and privacy

Other signals

  • security and privacy challenges in Gemini
  • agentic products
  • novel defense mechanisms
  • adversarial attacks
  • prompt injections
  • contextual security threats
  • autonomous and agentic AI systems