Postdoctoral Researcher - Multimodal AI for Whole-Body PET/CT

UniBE is a research-focused institution that values interdisciplinary collaboration. The role of a Postdoctoral Researcher in Multimodal AI for Whole-Body PET/CT involves developing advanced methods that combine PET/CT imaging with clinical text data. This includes creating context-aware image analysis and segmentation techniques, as well as multimodal representation learning and outcome prediction. The researcher will also explore the generation of physician-editable reports and lead projects from model design through multicentre clinical validation. The goal is to publish in leading international journals and present at conferences in the fields of machine learning and medical imaging. Additionally, the role involves contributing to grant applications and fostering interdisciplinary collaboration. The main responsibilities include shaping research questions, developing multimodal methods, and leading projects through clinical validation. The researcher will also aim to publish findings and present them at conferences, as well as contribute to grant applications and interdisciplinary efforts. The ideal candidate will have a PhD in computer science, machine learning, biomedical engineering, or a related field. Strong experience in deep-learning research on images, particularly with 3D/volumetric data, is essential. Excellent skills in Python and PyTorch, along with the ability to create reproducible pipelines at scale, are required. The candidate should have a track record of substantial methodological ownership and strong first-author publications. The ability to frame research questions, design baselines and ablations, identify leakage and shortcut learning, and communicate across disciplines is crucial. Prior experience with PET/CT is not required, as the institution provides the necessary clinical, biological, and imaging expertise. What the role asks for: - PhD in computer science, machine learning, biomedical engineering or a related field. - Strong deep-learning research on images. - Excellent Python/PyTorch skills. - Reproducible pipelines at scale. - Substantial methodological ownership. - Strong first-author publications. - Ability to frame research questions, design baselines/ablations, identify leakage and shortcut learning, and communicate across disciplines. - Nice-to-have: multimodal/vision-language learning, self-supervised pretraining or medical foundation models, longitudinal imaging, clinical NLP/report generation, computational pathology/whole-slide imaging, and distributed training.