Postdoctoral Fellow - Foundation Models for Spatial Biology

The Faisal Mahmood and Sizun Jiang Labs are seeking a postdoctoral fellow to develop foundation models for spatial proteomics and spatial transcriptomics. We are interested in learning how molecular profiles, cell states, and spatial relationships together explain tissue organization and disease. Our labs develop machine learning methods for computational pathology and multimodal biomedical data, including KRONOS for spatial proteomics and HEST and VORTEX for spatial transcriptomics research. This position will build on these efforts to develop models that generalize across tissues, patients, and measurement technologies. Research focus: Foundation model development: Design and train self-supervised models for multiplex protein imaging and spatial gene expression, learning representations at the molecular, cellular, and tissue levels. Multimodal learning: Integrate spatial proteomics, transcriptomics, and histopathology to connect tissue morphology with molecular function. Generalization: Develop approaches that accommodate variable marker and gene panels, missing modalities, and differences across platforms and patient cohorts. Biological discovery: Study cell states, tissue neighborhoods, and tumor and immune cell organization, with applications to biomarker discovery and treatment response research. The fellow will design rigorous benchmarks, collaborate with computational and experimental researchers and pathologists, lead publications, and contribute reproducible code and models. Required qualifications: A PhD, completed by the start date, in computer science, machine learning, computer vision, computational biology, bioinformatics, statistics, biomedical engineering, or a related field. Strong experience developing deep learning methods, demonstrated through publications, preprints, or substantial research software. Proficiency in Python and a deep learning framework such as PyTorch. Experience with representation learning, computer vision, multimodal learning, or machine learning for molecular data. Ability to formulate research questions, design careful experiments, and communicate findings clearly. Interest in understanding the biological meaning and limitations of model predictions. Preferred experience: Self-supervised pretraining, transformers, or foundation models. Distributed training and efficient processing of large imaging or molecular datasets. Spatial proteomics, spatial transcriptomics, single-cell analysis, or computational pathology. Modeling spatial relationships, integrating datasets, or evaluating generalization across independent cohorts. How to apply: Please email faisalmahmood@bwh.harvard.edu and sjiang3@bidmc.harvard.edu with the subject line 'Postdoc Application: Foundation Models for Spatial Biology' and include your CV, two or three representative publications or preprints, a brief research statement (up to two pages) describing your previous work and interests relevant to this position, and names and contact information for two professional references. Applications are reviewed on a rolling basis.