Research Assistant - PhD (m/f/d)

The Chair of Imaging and Machine Vision (LFB) conducts groundbreaking research at the interface of imaging instrumentation and computer-assisted imaging. Our spectrum ranges from the development of new hybrid imaging systems to advanced image reconstruction and video communication. Building on this broad interdisciplinary expertise, we focus on the robust 3D reconstruction of anatomical structures and multimodal image analysis. By combining hybrid machine learning with physical modelling, we develop tailor-made solutions that close the gap between raw image data and clinical application and thus ensure reliable decision support in complex medical workflows. Become part of the medical AI research as a member of a new interdisciplinary research consortium. We develop a fully automated process chain for surgical planning, with a specific focus on dysgnathia and extremity surgery. You work in a strong interdisciplinary team together with our consortial partners, including clinical experts from large university clinics (data & validation), specialised software partners (platform integration) and experts in manufacturing technology (production automation). The project aims to overcome the “data bottle neck” in the medical AI. Instead of relying exclusively on massive manual annotations, we want to use unlabeled data to build intelligent systems that truly “understand” 3D anatomy. We move from raw DICOM data to patient-specific, geometry-based surgical templates. The RWTH is certified as a family-friendly university. RWTH offers a variety of health, counselling and prevention services (e.g. university sports) as part of a university health management. There is a wide range of training and the possibility of obtaining a job ticket for employees and civil servants.

You are responsible for researching and developing next-generation AI methods for understanding medical 3D image data. The aim is to automate complex surgical planning processes using modern self-supervised techniques:

• Pioneering Self-Supervised Learning: Research and Adaptation of Modern Joint Embedding Predictive Architectures (e.g. I-JEPA, V-JEPA) for Volumetric Medical Data (CT/DVT). Their goal is to develop a “Volumetric-JEPA” that learns robust anatomical representations from unlabeled data without relying on pixel-level reconstruction.

• Data-Efficient Segmentation: Use of these pre-trained semantic backbones for highly precise segmentation of critical structures (e.g. jaw, nerve channels) by means of few-shot learning. This is intended to drastically reduce the need for manual annotations compared to traditional monitored methods.

• Generative vs. Predictive AI: Systematic comparison of the robustness of predictive architectures (JEPA) compared to generative approaches (e.g. diffusion models, GANs), in particular with regard to the handling of strong anatomical deformations and image artifacts (OOD-Robustheit).

• Multimodal VLM Interface: Development of interfaces to feed the learned high-level feature embeddings into Multimodal Large Language Models, creating a semantic “Safety Layer” that validates surgical plans based on anatomical understanding.

• Algorithmic Surgical Planning: Implementation of geometric algorithms (osteotomy levels, collision analyses) that build on the robust features of your AI models to automate the planning process.

• Education: Completed academic studies (master or comparable) in computer science, physics, engineering or a related field with strong focus on AI/ML.

• Technical skills: knowledge of Python and Deep Learning Frameworks (PyTorch). Experience with Self-Supervised Learning (SSL), Transformers or Modern Architectures such as Masked Autoencoders (MAE) / JEPA is expressly desired.

• Methodological knowledge: Strong understanding of computer vision, representation learning and high-dimensional geometry.

• Soft Skills: passion for solving medical challenges and the ability to work in multidisciplinary teams (engineers, clinicians).

The employment is in the employment relationship. The place shall be occupied at the next possible time and shall be limited to 3 years. The temporary employment is carried out within the framework of the time-limits of the science-time contract law. It is a full-time job. There is a doctoral opportunity. The grouping depends on the TV-L. The place is rated with EG 13 TV-L. The job description is aimed at all sexes. We want to promote the careers of women at RWTH Aachen University and are therefore looking forward to applicants. Women are preferably taken into account in the case of equal suitability, competence and professional performance, provided You are underrepresented in the organisational unit and, if not in the person of a competitor, outweigh the reasons. Applications for suitable people with difficulty are expressly desired. For the purposes of equal treatment, we ask you to waive an application photo. Information on the collection of personal data pursuant to Articles 13 and 14 of the General Data Protection Regulation (GDPR) can be found at https://www.rwth-aachen.de/dsgvo-information-bewerbe