Translational Post Doctoral Researcher - Agentic AI Neurodegeneration

Job Description

Johnson & Johnson Innovative Medicine is seeking a Translational Post Doctoral Researcher — Agentic AI for Neurodegeneration for a 2-year fixed-term position. This position can be located in either Raritan, NJ;
Titusville, NJ;
Spring House, PA;
San Diego, CA; or Cambridge, MA. No fully remote option.

The next frontier in neurodegeneration research is integrating insights across the data we already have at scale with agentic AI in ways that were previously not possible. Whole slide pathology, PET and MRI imaging, multi-omics, and longitudinal clinical records each offer a different lens on the neurodegenerative diseases; brought together, they tell a story no single modality can. This integration challenge is reshaping how we build agentic AI systems for drug discovery and how we evaluate them.

Traditional benchmarks were composed for single-modality reasoning. Evaluating whether an AI co-scientist can synthesize across pathology, imaging, molecular, and clinical evidence and produce hypotheses that are biologically sound, demands new frameworks. We are seeking a Postdoctoral Researcher to build them.

The Researcher will be embedded in the Machine Intelligence (MI) team at J&J Innovative Medicine, working in partnership with the c-brAIn academic network. The role begins with engagement in multi-modal neuroscience data – understanding what each modality reveals, how they relate, and where integration breaks down – and builds toward crafting the evaluation frameworks and standards by which agentic co-scientist systems are tested, validated, and trusted.

The Researcher will work day-to-day with AI scientists in J&J’s Machine Intelligence group while partnering closely with translational and experimental teams across C-BRAIN’s academic network at Washington University in St. Louis and partner institutions. Mentorship is designed to build leaders at the Multi-Modal Data × AI Evaluation × Neurodegeneration interface, with opportunities for publications, cross-sector exposure, and leadership development.

• Characterize and integrate biomedical data modalities – digital pathology (whole slide images), neuroimaging (PET, structural and functional MRI), omics (genomics, transcriptomics, proteomics, metabolomics), and longitudinal clinical data to develop specialized, domain-specific models for neurodegeneration
• Build and refine data engineering pipelines that harmonize heterogeneous modalities – reconciling differences in spatial resolution, temporal scale, and dimensionality – into unified analytical frameworks
• Identify where cross-modal integration produces genuine insight versus where it introduces noise or artifact, establishing ground truth for downstream AI evaluation

• Critically assess AI-driven literature synthesis and automated “third reviewer” capabilities for detecting methodological weaknesses, logical gaps, and unsupported claims across data modalities
• Establish standards for how agentic systems incorporate overlooked or contradictory evidence such as negative findings, failed clinical trials, etc. and evaluate whether these integrations generate genuinely novel hypotheses
• Design evaluation frameworks for agentic AI systems operating across neuroscience data modalities – assessing whether models can reason credibly across imaging, omics, and clinical evidence
• Develop benchmarks using synthetic and real-world multi-modal datasets that probe AI co-scientist capabilities under realistic research conditions, testing for robustness, reproducibility, and alignment with expert-level biomedical reasoning

• Serve as a neurodegeneration domain expert within the AI/ML team, ensuring that model outputs remain anchored to clinically relevant disease questions
• Translate evaluation findings into actionable guidance for AI system development, bridging computational and experimental perspectives
• Publish evaluation methodologies and findings in leading journals and conferences (e.g., AD/PD, AAIC, NeurIPS)
• Articulate emerging AI/ML approaches – causal reasoning, intent classification, agentic planning – to diverse…