Postdoc - AI Safety

About the role

Johnson & Johnson Innovative Medicine is recruiting a Postdoctoral Researcher in AI Safety for its Data, Data Science & AI organization. This is a fixed-term research appointment of two years reporting directly to the Scientific Fellow, AI Safety. The role can be based at one of the sites in Belgium, the Netherlands, the UK or Ireland. Agentic AI is central to pharmaceutical R&D, and the research focuses on governing it through provable controls, adversarial assurance and safety-by-design.

Key responsibilities

• Run an independent research agenda agreed with the Scientific Fellow on controls, adversarial assurance or safety-native architecture for agentic AI in regulated settings.
• Design rigorous, reproducible experiments (baselines, ablations, uncertainty estimates) to test whether safety properties hold.
• Evaluate pre-training data interventions and post-training methods such as supervised fine-tuning, preference optimization and safety tuning.
• Prototype layered architectures that constrain agent behavior and turn requirements into testable system specifications.
• Work with platform engineering to move validated methods into the GenAI Platform with reproducible, auditable documentation.
• Present work to scientific, engineering and leadership audiences and represent the team at conferences and standards activities.

Required profile

• PhD in computer science, AI/ML, applied mathematics or a closely related field, preferably completed before the start date.
• First-author research record with peer-reviewed publications or preprints in AI safety/alignment, AI/ML for cybersecurity, formal methods or a related area.
• Hands-on experience with foundation models and agentic AI (e.g. retrieval-augmented generation, tool use, planning, multi-agent frameworks).
• Excellent written and verbal English and scientific rigor in characterizing model and agent behavior.

Required skills

• Strong programming skills (e.g. Python and modern ML frameworks) and reproducible experiment building.
• Preferred: policy engines or authorization languages (OPA/Rego, Cedar), formal verification tools, fine-tuning foundation models for scientific/biomedical domains, privacy-preserving ML, open-source contributions to AI safety.

The anticipated pay range in the primary posting location is EUR 60,000 to EUR 96,255, with lower ranges for Ireland and the UK. Benefits include an annual bonus depending on performance.