Developing AI That Works for Healthcare

News

[Summer 2026] I will be joining the Hertie Institute for AI in Brain Health at the University of Tübingen as a Junior Group Leader, where I will start my own research group focused on interpretable foundation models for computational pathology funded by the BMFTR Zukunft eHealth program. I am actively looking for motivated people to join — see Open Positions below.

I am currently a Research Fellow in the lab of Prof Kun-Hsing Yu in the Department of Biomedical Informatics at Harvard Medical School in Boston. Previously, I completed my PhD in Medical Engineering and Medical Physics in the Harvard-MIT Division of Health Sciences and Technology at MIT under the guidance of Prof Jayashree Kalpathy-Cramer. From 2023 to 2024, I was a Research Fellow in the Department of Data Science at Dana-Farber Cancer Institute in Boston. I also hold an MD from Heidelberg University Medical School and a BS in Physics with a minor in Computer Science from Kiel University in Germany.

My research focuses on developing AI methods that are clinically meaningful, robust, and aligned with real-world healthcare workflows. I am particularly interested in combining technical innovation with domain expertise so that models can support clinicians and improve patient outcomes.

Research Areas

I am currently building a research program around three connected themes:

1. Disease modeling from complex clinical and molecular data
   Building machine learning models that can represent biological and clinical heterogeneity in diseases such as cancer.
2. Interpretability and trust in medical AI
   Designing methods that help clinicians understand model behavior and identify when predictions are reliable.
3. Translation to healthcare practice
   Creating AI systems that are tailored to clinical needs, fit existing workflows, and can be evaluated for practical impact.

Detailed research pages are in development and will be added here soon.

For a full list of publications, please see my Google Scholar profile.

CV and Contact

You can download my current CV here: Download my CV.

I am always happy to connect about potential collaborations in medical AI, particularly in the oncology and pathology space.

Open Positions

I am recruiting for my new group at the Hertie Institute for AI in Brain Health, University of Tübingen, launching summer 2026. We are a collaborative and interdisciplinary group developing transparent and trustworthy AI for medical image analysis. No prior experience with medical data is required — openness to interdisciplinary research is what matters most.

Informal inquiries are welcome before submitting a formal application — feel free to email me at kathi@alum.mit.edu with the subject line “pathologically curious”.

Postdoc in Interpretable Computational Pathology

Full-time · 5 years · Start: July 1, 2026 · Deadline: June 17, 2026 · Official job ad

You will contribute to the implementation of a funded research agenda focused on multimodal histopathology (specifically glioma diagnosis), while maintaining the creative freedom to refine methodologies that bridge the gap between complex deep learning and clinical utility.

Tasks

• Methodological lead for deep learning model development for histopathology and clinical data
• Act as the primary technical bridge between engineering colleagues and medical professionals
• Prepare high-impact publications and present at national and international conferences
• Support supervision of graduate students and contribute to grant writing

Requirements

• PhD in a computational field (Computer Science, Data Science, Biomedical Informatics, or related)
• Strong programming skills (Python or R) and experience with large-scale ML pipelines
• Track record of scientific rigor evidenced by first-authored publications
• Excellent written and spoken English — German is not required

Preferred

• Experience with digital pathology or other clinical imaging modalities
• Experience integrating imaging data with clinical or genomic modalities
• Self-starter mindset comfortable helping shape workflows in a newly established lab

PhD in Trustworthy AI for Medical Image Analysis

75% · 4 years · Start: July 1, 2026 · Deadline: June 17, 2026 · Official job ad

This PhD project focuses on the interpretability of foundation models for automated glioma diagnosis and subtyping, aiming to make these “black-box” systems transparent for clinical use by quantifying biological tumor characteristics.

Tasks

• Train and refine deep learning models for automated glioma diagnostics with a focus on interpretability
• Develop novel methods to quantify biological characteristics of tumors and validate against clinical benchmarks
• Present research at internal and international forums and prepare manuscripts for publication
• Participate in grant writing and collaborative research as you grow toward scientific independence

Requirements

• Master’s degree (or equivalent) in a quantitative field (e.g., Computer Science, Physics, Data Science, Biomedical Engineering, or Neuroscience)
• Programming skills (Python or R) and prior experience with ML/DL applications
• High degree of independent thinking and curiosity
• Functional command of English — German is not required

Preferred

• Interest in oncology or neuropathology (prior medical knowledge not required)
• Desire to work closely with clinicians and engineers in a multidisciplinary team
• Initial experience or interest in academic writing

Master & Bachelor Students

I have multiple rotation and thesis projects available in interpretable ML for histopathology. Topics are flexible — potential directions include agentic systems to quantify pathologic concepts, interpretable attention maps for glioma diagnosis, and more. I am also happy to discuss your own ideas.

To get in touch, email kathi@alum.mit.edu with the subject line “pathologically curious” and a few words about your research interests.