Postdoctoral Researcher in Video Understanding
Johnson & Johnson Innovative Medicine
Before joining Johnson & Johnson in September 2026, I was a postdoctoral researcher in the Krauthammer Lab at the Department of Quantitative Biomedicine, University of Zurich, where I worked on generative modeling for protein design and biomedical imaging.
I did my PhD at the Institute of Computer Graphics and Vision, Graz University of Technology, supervised by Prof. Thomas Pock, with Prof. Carola-Bibiane Schönlieb as external referee. I defended my thesis "Variational Methods in Imaging Meet Machine Learning" in June 2024. During my PhD I worked on inverse problems in medical imaging, bilevel optimization, and generative modeling using diffusion models and flow matching for image segmentation.
I'm interested in generative modeling (flow matching, diffusion and energy-based models) and its applications in biomedicine, ranging from video understanding and cell morphology to protein fitness optimization and medical image analysis. I also have a strong background in variational methods and inverse problems in imaging.
* indicates equal contribution.
ECCV-W 2026 ECCV Workshop on BioImage Computing (BIC)
We introduce a joint flow matching approach for continuous, dose-conditioned prediction of cell morphology under compound treatment, rather than treating drug concentrations as discrete categories.
NeurIPS 2026 Conference on Neural Information Processing Systems
We present EBMol, an energy-based model that learns atom-additive potentials for 3D molecule generation, ensuring physical consistency through the Boltzmann distribution and achieving state-of-the-art results on QM9 and GEOM-Drugs.
arXiv 2026 arXiv preprint
We present CHASE, a framework for direct high-fitness protein variant generation by training a conditional flow-matching model on compressed protein language model embeddings within a compact latent space.
Nat. Commun. 2026 Nature Communications
CRISPR-PAMdb, a large-scale database of Cas9 proteins and PAM profiles, is introduced alongside CICERO, a protein language model–based predictor that enables accurate PAM preference inference and broad exploration of PAM diversity for genome editing.
CVPR 2026 IEEE/CVF Conference on Computer Vision and Pattern Recognition
We propose an improved MeanFlow training strategy that rapidly stabilizes instantaneous velocity before progressively emphasizing long-interval averages, enabling faster convergence and higher-quality one-step generation.
WACV 2026 IEEE/CVF Winter Conference on Applications of Computer Vision
We introduce Restora-Flow, a training-free method for mask-based inverse problems that guides flow matching sampling by a degradation mask and incorporates a trajectory correction mechanism to enforce consistency with degraded inputs.
NeurIPS 2025 Conference on Neural Information Processing Systems
An energy matching framework is introduced, combining optimal transport paths far from the data manifold with an entropic energy term to explicitly capture data likelihood, enabling flexible priors and high-fidelity generation without auxiliary networks.
ICML 2025 International Conference on Machine Learning
A variational framework for generative protein fitness optimization using a flow matching prior and a classifier-guidance model in a continuous latent space.
IJCV 2025 International Journal of Computer Vision
A flow matching framework for generative medical image segmentation using signed distance functions that enables canonical smoothing of SDF mask distributions through noise injection.
JBHI 2025 IEEE Journal of Biomedical and Health Informatics
A multi-channel generative approach based on flow matching synthesizes medical images paired with heatmaps, enabling robust data augmentation for anatomical landmark localization, especially with limited training data or occlusions.
JMIV 2025 Journal of Mathematical Imaging and Vision
A bilevel learning framework for variational image reconstruction that uses a primal-dual approach with a-posteriori error bounds and adaptive step sizes.
SASHIMI 2024 Simulation and Synthesis in Medical Imaging (MICCAI Workshop)
A DDPM-based framework for generating medical images with landmark heatmaps, using a Markov Random Field for matching and a Statistical Shape Model for plausibility checks.
GCPR 2023 DAGM German Conference on Pattern Recognition Honorable Mention
A score-based generative model for binary medical image segmentation using signed distance functions, where diffusion corrupts SDF masks instead of binary ones for more natural distortions.
SSVM 2023 International Conference on Scale Space and Variational Methods in Computer Vision
A framework for incorporating general discretizations of second-order TGV with variational consistency, learning interpolation filters via a piggyback algorithm.
HCB 2022 Histochemistry and Cell Biology
A joint alignment and reconstruction algorithm for electron tomography without fiducial markers, applied to studying immune–beta cell interactions in NOD mice for type 1 diabetes research.
SIMODS 2022 SIAM Journal on Mathematics of Data Science
An analysis of a piggyback-style method for differentiating saddle-point problems, applied to learning optimized shearlet transforms for imaging.
OAGM 2021 Computer Vision and Pattern Analysis Across Domains
Deep learning is used to study insulin granules in NOD mouse beta cells, where a multi-task regression aids in distinguishing healthy from diabetic samples.
OAGM 2021 Computer Vision and Pattern Analysis Across Domains
Deep neural networks can estimate grain density in austenitic steel, with classification and regression models learning distinct feature representations.