Director, Model & Agentic Learning

About the Role

Johnson & Johnson Innovative Medicine is recruiting a Director, World Model & Agentic Learning to join its Data, Data Science & AI organization. This newly created leadership role reports directly to the Head of Generative AI and will lead the AI science team that builds our enterprise world model and agentic-learning capability for the R&D agentic AI platform.

• World Model Design – Design how agents represent accumulated domain understanding and reason against it, rather than re‑deriving knowledge from raw sources on each task; build mechanisms to represent the system’s confidence, boundaries, gaps, and contradictions; ensure knowledge earned in one domain compounds and surfaces wherever relevant; serve the representation as queryable, grounded knowledge with provenance and confidence; curate the system’s proposals by validating, deduplicating, and resolving conflicts.
• Agentic Learning Design – Design the mechanisms that turn operation into improvement (e.g., active learning from expert corrections, memory‑based or in‑context learning, outcome‑driven refinement); make every run, expert correction, and decision outcome a signal that improves the next result; keep institutional understanding fresh and honest as sources, evidence, and experts change over time.
• Expert Partnership – Partner with scientists and domain experts so their expertise becomes something the system can apply consistently at scale; maintain and apply experts’ judgment without overriding it.
• Accountability & Evaluation – Define and prove the accountability bar: demonstrate that the system produces better decisions over time; make every conclusion auditable and reconstructable; partner with J&J Technology, Generative AI evaluation, and AI operations teams to consume per‑decision outcome signals and validate decision‑quality improvement rigorously.
• Team Leadership – Recruit, build, and lead a team of 4–8 AI scientists; attract, develop, and retain top talent in continual learning, knowledge representation, and agentic systems; establish a culture of scientific rigor, ownership, and accountability within the team.

• Generation‑first role: the hard problem is knowledge accumulation and learning over time, not content generation.
• Platform or application engineering: this role builds on the existing platform and does not own deployment or engineering.
• Evaluation governance: the Generative AI evaluation function owns independent evaluation; this role partners with it.
• The data or memory substrate: the platform’s governed data and context/memory layers manage data and orchestration; this role references and builds on them.

• Minimum – 8+ years post‑academic industry experience building and shipping AI/ML systems; significant technical architecture ownership; deep, hands‑on expertise with modern AI systems (LLMs, retrieval‑augmented generation, agentic frameworks, knowledge representation); experience designing systems where knowledge accumulation, memory, or continual learning is central; experience learning from real‑world operation and expert feedback (e.g., active learning, in‑context learning, outcome‑driven refinement); strong people‑leadership experience (recruiting, building, leading technical or scientific teams);
  
  ability to set and defend a technical architecture; excellent communication skills.
• Preferred – PhD in computer science, AI/ML, applied mathematics, computational science or related discipline; experience in regulated or high‑stakes environments (e.g., life sciences, healthcare, finance); background in life sciences, drug discovery, or pharmaceutical R&D; experience with knowledge graphs, ontologies, or explicit knowledge representations; track record building auditable, traceable AI systems; publications or recognized contributions in continual learning, agentic systems, knowledge representation, or human‑in‑the‑loop AI;
  
  experience partnering with enterprise platform and IT delivery organizations; experience building reusable frameworks or platform capabilities; experience defining clean interfaces between knowledge/memory substrate and reasoning or…