Senior Security Engineer (AI Safety), London, Lausanne

2 months ago

London, United KingdomSenior

Responsibilities

Conduct threat modeling for AI/ML vulnerabilities and develop an AI-specific risk management framework.
Inventory, classify, and protect ML model weights, code, and training data.
Design and deploy LLM guardrails, sandboxing, monitoring, and security controls for ADK, MCP, and autonomous agentic workflows.
Implement security controls across data ingestion, training pipelines, and inference runtimes in collaboration with ML researchers and platform engineers.
Lead AI security incident response and use machine learning for threat hunting, anomaly detection, and mitigation.
Automate AI risk posture monitoring and continuous compliance metrics for emerging regulations including the EU AI Act.
Implement AI Safety standards in coordination with Legal, Compliance, Google DeepMind, and external partners.

Deep practical and conceptual understanding of deep learning frameworks, large-scale cloud training or inference infrastructure, and LLM ecosystems.
Familiarity with AI security threats including prompt injection, model inversion, and data poisoning, plus frameworks such as OWASP Top 10 for LLMs or MITRE ATLAS.
Proven experience securing agentic frameworks such as ADK and MCP and implementing agent-to-agent identity and access controls.
Proficiency in cloud platform security, preferably GCP, including container security, multi-cloud or SaaS integrations, and network isolation.
Ability to define pragmatic security mitigations in a high-velocity scientific research environment.
Ability to write production-grade code, preferably Python, for security tooling, telemetry collection, and policy enforcement.
Strong communication skills for translating machine learning risks into actionable engineering work.
Preferred experience with AI red teaming, vulnerability research, or simulated attacks against LLMs, agent networks, and ML backends.
Preferred exposure to biotech, pharma, or deep-tech regulated environments and GxP compliance.
A BSc, MSc, or PhD in Computer Science, Machine Learning, Cybersecurity, or a related quantitative field is preferred.
Relevant security or cloud credentials such as OSCP or Professional Cloud Security Engineer are preferred.

Hybrid working requires attendance in the office three days per week, currently Tuesday, Wednesday, and one additional team-dependent day.
Accommodation discussions are available for candidates unable to follow the standard hybrid approach.
The company offers an interdisciplinary, collaborative culture focused on learning, shared knowledge, and meaningful scientific impact.

201-500 employees