Senior Scientist - AI Safety

About the role

Johnson & Johnson Innovative Medicine is recruiting a Senior Scientist, AI Safety to join its Data, Data Science & AI organization in Madrid or Barcelona. This is a newly created scientific role reporting directly to the Scientific Fellow, AI Safety, within a hybrid model of 3 days per week in the office. The role defines how agentic AI used in pharmaceutical R&D can be governed through provable controls and continuously tested through adversarial assurance, then shapes safety-native AI architectures.

Key responsibilities

• Design machine-readable control policies governing agent actions, tool use, data access and information flow, and implement deterministic policy enforcement with auditable decisions and human-approval paths.
• Build continuous red-teaming methods for agentic AI and embed adversarial evaluation in the GenAI Platform to test models, agents, tools and workflows.
• Research and prototype safety-native architectures, including pre-training data interventions and post-training methods such as supervised fine-tuning, preference optimization and safety tuning.
• Translate scientific, quality, privacy and regulatory requirements into testable system specifications.
• Define accountability measures and ensure control decisions are traceable as audit evidence for scientific, quality and regulatory review.
• Partner with R&D scientists, platform engineering and security, privacy, legal and quality functions, and represent the work externally through publications and standards contributions.

Required profile

• PhD in computer science, AI/ML, applied mathematics or a closely related technical field (required).
• At least one year of post-PhD research or industry experience developing and deploying AI/ML, autonomous-agent or security-critical systems.
• Deep hands-on expertise in agentic AI, including foundation models, retrieval-augmented generation, tool orchestration, memory, planning and multi-agent frameworks.
• Demonstrated research or applied expertise in AI safety and alignment, such as supervised fine-tuning, preference optimization, safety tuning, adversarial evaluation or interpretability.

Required skills

• Expertise in at least one of: policy-as-code and authorization, adversarial machine learning, formal methods or provenance systems.
• Experience adversarially evaluating deployed AI systems and detecting evaluation leakage, overfitting and weak baselines.