Selected work.

Faster models do not automatically mean faster science. Experiments take time, infrastructure matters, and researchers still need to assess what a result means.

Our paper in Scientific Reports brings together a scoping literature review, a framework for assessing AI capabilities across the research lifecycle, and expert elicitation with eight biomedical researchers. It examines where acceleration is plausible and what could limit it.

I co-authored the work in the Samwald Lab at the Medical University of Vienna. The question continues to shape how I think about building AI for science: which change would actually help a lab learn faster?

Understanding how a limb regenerates requires measuring how its tissues change. Doing that manually across microscopy volumes containing roughly a thousand slices is slow and difficult to reproduce.

I am building and evaluating convolutional models to segment the epidermis and stratum compactum in HREM images of regenerating axolotl limbs. The aim is to turn these image volumes into reproducible 3D tissue measurements.

The work is in progress. My contribution spans the image-analysis workflow, model development, and evaluation of the resulting segmentations.

I founded The Residency Vienna, a 500 m² hacker house where founders live and work on their own companies.

The useful part is the everyday proximity: sharing meals, talking through a problem, and seeing someone else keep going with theirs. It is also a home. I think of that shared life as a modern form of extended family.