Senior Research Scientist, AI Safety and Security

Google Google · Big Tech · Singapore

Research Scientist focused on AI Safety and Security, specializing in proactive threat mitigation, adversarial machine learning, and agentic security. The role involves foundational research in model robustness, continual learning, interpretability, and developing advanced evaluation benchmarks for trustworthy AI.

What you'd actually do

  1. Drive foundational machine learning research in model robustness, continual learning, interpretability, and multiobjective optimization to advance trustworthy AI.
  2. Design and develop rigorous evaluation protocols, scenario-based benchmarks, and stress-testing methodologies to assess frontier AI capabilities and multi-agent consensus.
  3. Curate advanced datasets and conduct fine-tuning or optimization experiments to enhance model resilience against emerging threats and ensure adherence to safety constraints.
  4. Collaborate extensively with regional engineering hubs, core product teams, and academic partners to transition theoretical proofs-of-concept into robust production solutions.
  5. Publish groundbreaking research in machine learning venues and actively participate in academic and industry research communities.

Skills

Required

  • Machine learning
  • Adversarial machine learning
  • Evaluating frontier AI systems
  • Supervised learning
  • Unsupervised learning
  • Reinforcement learning
  • ML interpretability
  • Adversarial robustness
  • ML safety
  • Generative models
  • Agentic AI
  • Multi-object optimization
  • Scientific publication submission(s)
  • General purpose programming languages (e.g., Python)
  • Investigating emerging technical threats
  • Designing robust, proactive defense mechanisms
  • AI agent security
  • Data poisoning
  • Prompt injection
  • Model backdoor detection
  • Debugging complex ML failure modes
  • Reverse engineering model behaviors
  • Red-teaming frontier AI systems

Nice to have

  • Continual learning
  • Multiobjective optimization
  • Scenario-based benchmarks
  • Stress-testing methodologies
  • Multi-agent consensus
  • Advanced datasets curation
  • Fine-tuning
  • Optimization experiments
  • Safety constraints
  • First-authored publications

What the JD emphasized

  • PhD degree in Computer Science, a related field, or equivalent practical experience.
  • 2 years of experience leading a research agenda.
  • Experience in machine learning, adversarial machine learning or evaluating frontier AI systems, which includes but not limited to supervised learning, unsupervised learning and reinforcement learning, ML interpretability, adversarial robustness, ML safety, generative models, agentic AI, multi-object optimization.
  • One of more scientific publication submission(s) for conferences, journals, or public repositories (such as CVPR, ICCV, NeurIPS, ICML, ICLR, etc.).
  • Demonstrated expertise in adversarial machine learning, AI agent security, data poisoning, prompt injection, and model backdoor detection.
  • Strong background in applying a security mindset to artificial intelligence, including debugging complex ML failure modes, reverse engineering model behaviors, and red-teaming frontier AI systems.
  • First-authored publications in top machine learning, safety/security tracks in machine learning or AI conferences, or HCI conferences.

Other signals

  • AI Safety
  • Adversarial ML
  • Interpretability
  • Evaluation Benchmarks
  • Trustworthy AI