PhD position Organ-Specific Embedding Models for Computational Pathology

As a PhD candidate, you will establish and benchmark self-supervised learning (SSL) pipelines tailored to specific organs, enabling accurate whole slide image (WSI) classification and downstream predictive tasks. You will lead the design, implementation, and evaluation of novel organ-specific embedding models — experimenting with contrastive objectives, innovative embedding strategies, and feature aggregation methods — and validate their performance in clinically relevant applications. The explainability of the learned models will also be thoroughly investigated.

Would you like to know more about the different phases within the PhD trajectory? You can read more about this on this page.

• Innovative Model Research: Formulate and test self-supervised objectives (e.g., DINOv2, Barlow Twins) and model architectures, exploring novel approaches to capture organ-specific histological features;
• Experimental Design & Analysis: Lead hypothesis-driven experiments, define evaluation metrics, conduct rigorous ablation studies, and interpret model behaviour to uncover biological insights from WSIs;
• Dataset Creation: Collaborate with IT and data management teams to centralize and index thousands of WSIs across organ groups, maintaining a robust database with redundancy and secure backups;
• Cross-Organ Benchmarking: Systematically benchmark your models against general-purpose foundational models.
• Interdisciplinary Collaboration: Work closely with clinicians, molecular pathologists, and bioinformaticians to integrate AI models into clinical practice;
• Scholarly Dissemination: Author high-impact publications, present at top-tier conferences, and contribute to grant and project proposals that extend your research vision.

You are a proactive, analytical thinker with a strong background in computer science, machine learning, or a related discipline, and an interest in medical AI. You thrive in an interdisciplinary environment and are eager to push the boundaries of computational pathology.

You have:

• A Master’s degree in Computer Science, Artificial Intelligence, Biomedical Engineering, or a related field (with outstanding grades);
• Hands-on experience with Python and deep learning frameworks (e.g., PyTorch, TensorFlow);
• Familiarity with self-supervised learning techniques and multiple instance learning, or a strong willingness to learn. A strong mathematical background is preferred;
• Knowledge of containerization (Docker/Kubernetes), distributed computing, and version control (Git), or a willingness to learn;
• Excellent communication skills and the ability to collaborate effectively with researchers, clinicians, and IT specialists.