
Post Doc - AI Safety
AI
Postdoctoral Researcher, AI Safety at Johnson & Johnson Innovative Medicine, joining the Data, Data Science & AI organization for a fixed-term two-year research appointment. The role leads a focused, publishable research program on governing agentic AI in pharmaceutical R&D through provable controls, adversarial assurance, and safety-native architecture. Requires a PhD in computer science, AI/ML, applied mathematics, or a closely related technical field, with first-author research in AI safety/alignment or related areas and hands-on experience with foundation models and agentic AI.
What you'll do
- Execute an independent research agenda 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.
What you need
Must-have
PythonMachine Learning frameworksFoundation modelsAgentic AIRetrieval-Augmented GenerationTool usePlanningMulti-agent frameworks
Nice-to-have
OPA/RegoCedarFormal verification toolsPolicy enginesAuthorization languagesPrivacy-preserving MLData de-identificationData provenance
Soft skills
Excellent written and verbal communicationScientific rigorAbility to present technically defensible arguments
PostdocAI SafetyPharmaceutical R&DAgentic AIAI AlignmentAdversarial AssuranceMachine LearningResearchJohnson & Johnson