Machine Learning Engineer – Medical Imaging AI (Informatiker/in)

Germany | Sept. 3, 2026

Report as Closed

Company: Universitätsklinikum Frankfurt

Country: Germany

Type: Onsite

Employment: Full-time

Description: ...are you right in the middle

You don't want your models to end up in a publication, but rather arrive in the clinic?

Then you've come to the right place. We are looking for machine learning engineers for the Machine Learning working group at the University Medical Center Frankfurt - people who not only find medical AI exciting, but also drive it forward with conviction.

The focus is on the BMBF-funded project OMI (Open Medical Inference), which is developing an open platform as part of the medical informatics initiative to make AI models usable across locations and interoperably in clinical radiology - via open protocols and standards (FHIR, DICOMweb), without any Clinic has to operate models itself. They develop production-ready AI models for medical imaging that arrive exactly where they are needed.

In addition, we work in the RACOON network (https://racoon.network) with 38 university hospitals. You will take responsibility for end-to-end ML pipelines and have the freedom to develop solutions that can evolve into competitive AI applications. What is more important to us than the number of years you have worked is versatility, self-drive and the willingness to tackle problems head-on.

Whether you have just graduated and want to build your first productive pipelines or have already had end-to-end responsibility: If you work independently, deliver reliably and familiarize yourself with new topics, you will find the space to grow quickly with us. Experienced team members and a close connection to clinical research ensure that you are not alone. The position is initially limited to 12 months. A doctorate is possible by individual arrangement.



Your tasks:

End-to-end ML development

- Development, training, validation and deployment of deep learning models for medical image analysis (primarily CT and MRI)
- Work with the current spectrum of methods - from 2D/3D CNNs and the U-Net family to Transformer architectures on foundation models and vision language approaches; You choose the right tool for the respective clinical question
- Adaptation and fine-tuning of pre-trained foundation models for segmentation and classification, instead of training each model from scratch
- Responsibility for the entire pipeline: from DICOM data acquisition to containerized deployment - to the extent you want to undertake
- Rapid prototyping: the quick path from clinical problem to working prototype
- Performance optimization for accuracy, inference speed and Compute efficiency

Infrastructure, MLOps & platform development

- Setting up and operating training environments: configuring GPU servers and managing compute resources
- Building reusable ML infrastructure: modular pipelines, data loaders, training frameworks
- Reproducible workflows with modern MLOps tooling: experiment tracking (e.g. MLflow, Weights & Biases), data and model versioning (e.g. DVC), orchestration (e.g. Kubeflow, Airflow, Prefect)
- Deployment systems: Inference Services (e.g. FastAPI, KServe/BentoML), APIs and PACS integration, containerized via Docker/Kubernetes
- Monitoring of models in operation, including detection of data and model drift

Medical Data Engineering & Compliance

- DICOM/FHIR data processing: robust parsers, converters and quality control systems
- Data versioning and lineage for reproducible experiments
- Privacy-preserving workflows compliant with GDPR, German health data protection law (GDNG) and the requirements of the EU AI Act
- Traceability, audit trails and human-in-the-loop principles for clinical use Models
- Multi-institutional data integration across the 38 RACOON partner institutions

Research contribution

- Publication in top venues for medical AI (RSNA, ECR, MICCAI, Medical Image Analysis)
- Participation in collaborative research projects (RACOON, COMPARE, ENRICH)
- Doctorate possible by individual arrangement

...your profile is in demand

We evaluate applications based on potential and drive, not based on a checklist. Even if you don't meet every point, but enjoy getting involved: Apply.

Education & technical background

- Completed studies (Bachelor's or Master's) in (medical) informatics, physics, mathematics, data science or comparable
- Solid machine learning knowledge and experience in training neural networks - from studies, your own projects or job
- Secure Python programming: PyTorch or TensorFlow, NumPy, scikit-learn, pandas

Hands-on skills

- Experience in training CNNs, vision transformers or comparable architectures - whether during studies, internships, own projects or professionally
- Basic Linux knowledge (or clear willingness to learn): shell scripting, SSH, file systems, process management
- Familiarity with Git workflows; Code review practices and CI/CD fundamentals are a plus

Ideal, but not required

- Experience with foundation models and prompt-based segmentation (e.g. SAM/MedSAM2, SAM-Med3D, VISTA3D, nnU-Net)
- Practice with MLOps tooling (MLflow, W&B, DVC, Kubeflow) and containerized deployment (Docker, Kubernetes)
- Familiarity with the PyTorch ecosystem, MONAI or Hugging Face
- Knowledge of 3D/volumetric image processing and domain generalization across scanners and locations
- Understanding of regulatory frameworks for clinical AI (EU AI Act, MDR)

What sets you apart

- Self-drive: You identify problems yourself, develop solutions and implement them - even without Specification
- Versatility and hands-on mentality: You familiarize yourself with unknown topics and get stuck into it
- Reliability: You can rely on your promises
- Pragmatism: You combine research quality with the aim of delivering functioning systems
- Genuine interest in AI in healthcare as a driving force for your work
- Very good knowledge of German and good English

For experienced candidates: If you already have end-to-end pipelines If you have been responsible for setting up ML infrastructure, adapting foundation models or publishing them in top venues, the position offers space for a senior role with appropriate scope for creativity and mentoring tasks.

- Due to legal regulations, valid proof of measles immunity / measles vaccination is necessary.

...you have a lot on offer

- Collective agreement TV-UKF
- 30 days vacation, 38.5 hours / week, annual special payment, company pension plan
- Free state ticket for Hesse
- Real ownership: your models arrive in clinical workflows, not in the drawer
- Work on the current state of research: foundation models, multicenter data, real clinical translation
- Structured induction and mentoring by experienced team members
- University clinic campus, cafeteria, cafés
- Work-life balance, Part-time opportunities
- Health promotion
- Corporate benefits: Discounts & discounts on popular brands for our employees
- Daycare places, holiday care (information from family service)
- Insights: Instagram, YouTube, LinkedIn
- FAQs for new employees

Become part of our team now!

Contact: Dr. Andreas Bucher
Mail:bucher@med.uni-frankfurt.de**
**Application deadline: **September 25, 2026
**

Required documents: CV, cover letter including motivation (and relevant previous experience, if applicable), references and certificates from Degrees and, if necessary, additional qualifications

We are addressing this advertisement to applicants of all genders. Women are underrepresented in these positions at the University Hospital Frankfurt. Therefore, applications from women are particularly welcome. Severely disabled applicants will be given priority if they have the same personal and professional suitability.

Apply here:

Web: Apply here

Emails: