Postdoctoral Researcher, Ai Safety in Maidenhead

About the Role
Johnson & Johnson Innovative Medicine is recruiting a Postdoctoral Researcher, AI Safety to join our 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 our sites in Belgium, The Netherlands, UK or Ireland
Agentic AI is becoming central to pharmaceutical R&D—from discovery and translational science to development and regulatory work—where evidence standards are rigorous and errors can ultimately affect patient safety and outcomes. Our GenAI Platform supports that shift across a rapidly expanding population of autonomous workflows. Safety at this scale cannot be retrofitted through checks written into individual applications; it must be a property of how these systems are built.
As a Postdoctoral Fellow, you will lead a focused, publishable research program on how agentic AI in pharmaceutical R&D can be governed through provable controls, continuously tested through adversarial assurance, and made safe by construction. Your research agenda will sit within one or more of the team's three connected mandates:
Provable controls. Develop deterministic, explainable controls that persist throughout agentic workflows, and methods to verify that they hold.
Adversarial assurance. Advance continuous red-teaming and evaluation methods that test safeguards against credible failure scenarios and produce defensible evidence.
Safety-native architecture. Investigate pre- and post-training safety alignment and defense-in-depth techniques that make agentic AI safe by construction for regulated pharmaceutical R&D.
This is a hands-on research role with real systems as the testbed. You will formulate research questions, build the prototypes and experiments that answer them, publish the results, and work with our engineering partners to carry validated methods into the platform.

Key Responsibilities

Research Program

• Execute an independent research agenda (agreed with the Scientific Fellow/mentor) on controls, adversarial assurance, or safety-native architecture for agentic AI in regulated scientific settings.
• Design rigorous, reproducible experiments—including baselines, ablations, and uncertainty estimates—that test whether a safety property genuinely holds.
• Evaluate pre-training data interventions and post-training methods—including supervised fine-tuning, preference optimization, and safety tuning—for regulated scientific use cases, including whether safety properties persist under domain adaptation.
• Prototype layered architectures that constrain agent behavior, and translate scientific, quality, privacy, and regulatory requirements into testable system specifications.
• Work with platform engineering to move validated methods from prototype into the GenAI Platform, with documentation that makes results reproducible and auditable.
• Present your work to scientific, engineering, and leadership audiences, and represent the team at conferences, workshops, and standards activities.

What This Role Is Not

• Not frontier model development. We are not pre-training foundation models at scale. The research question is how alignment and architecture should be adapted so that the models and platforms available to us are safe for pharmaceutical R&D.
• Not a guardrail-prompt role. Safety here is architectural and enforcement is deterministic. A system prompt asking a model to behave is not a control.
• Not research in isolation. Your research questions come from real agentic workflows, and success includes evidence that the methods work on them—not only a publication.
• Not a production engineering or operations role. You prototype and validate; platform teams own production deployment, on-call support, and long-term maintenance.

Key Qualifications

• PhD in computer science, AI/ML, applied mathematics, or a closely related technical field, preferably completed before the start date.
• A record of first-author research, evidenced by peer-reviewed publications or preprints, in AI safety/alignment, AI/ML for cybersecurity, formal methods, or a closely related area.
• Hands‑on experience with foundation models and agentic AI, such as retrieval-augmented generation, tool use, planning, or multi‑agent frameworks, and their failure modes.
• Strong programming skills (for example, Python and modern ML frameworks) and experience building reproducible experiments.
• Excellent written and verbal communication in English, with the ability to present technically defensible arguments to scientific and engineering audiences.
• Scientific rigor in characterizing model and agent behavior, including uncertainty, limitations, failure conditions, and the strength of supporting evidence.

Preferred Qualifications

• Experience with policy engines, authorization languages (for example, OPA/Rego or Cedar), or formal verification tools.
• Experience fine‑tuning or evaluating foundation models for scientific or biomedical domains.
• Exposure to privacy‑preserving ML, data de‑identification, or data provenance, including their limitations.
• Interest in life sciences or another regulated, high‑stakes domain
• Open‑source contributions to AI safety, evaluation, or security tooling.

Location

This position will be located at one of our offices in: Beerse, Belgium; Spring House, Pennsylvania, United States; Raritan, New Jersey, United States. Hybrid work arrangements apply.

Johnson & Johnson is an Equal Opportunity Employer. All qualified applicants will receive consideration for employment without regard to race, color, religion, sex, sexual orientation, gender identity, age, national origin, disability, protected veteran status or other characteristics protected by federal, state or local law. We actively seek qualified candidates who are protected veterans and individuals with disabilities as defined under VEVRAA and Section 503 of the Rehabilitation Act.

Johnson & Johnson is committed to providing an interview process that is inclusive of our applicants' needs. If you are an individual with a disability and would like to request an accommodation, please contact us via Apply via the company portal.

#J-18808-Ljbffr