At HeiGIT, I built a reproducible pipeline that processed more than 105 million street-view image sequences into global road surface datasets, with continental F1 scores of 91–97%. I also co-authored a peer-reviewed paper and completed my M.Sc. Geography thesis on classifying road surface pavedness across 4 million km of roads.
At the Max Planck Institute for Astronomy, I built deep-learning pipelines to reconstruct astronomical imagery and automated workflows to evaluate photometric accuracy. My work brings together geospatial data, machine learning, and reproducible processing, including open data published to the Humanitarian Data Exchange.

