I am a research scientist at Google DeepMind with a Ph.D. in deep learning, currently leading efforts on medical AI agents and watermarking. I co-created and deployed SynthID, Google’s watermarking solution. I co-developed Med-Gemini, a family of Gemini-based medical LLMs, and AMIE, a diagnostic dialogue AI. I also contributed to safety evaluations in Bard (now Gemini). I successfully initiated and led multiple cross-organizational research projects , resulting in several deployments and Nature submissions. I co-authored numerous papers, and my open-source projects are widely adopted. My work is frequently covered in public media and I maintain a blog here.

My research expertise spans:

• Deep learning: robust deep learning, watermarking models, diffusion models, variational auto-encoders, model quantization, and distillation.
• AI agents: dialogue agents, multi-agent systems, auto-raters, human evaluation, and reasoning agents.
• AI for science: health, incl. medical imaging, medical LLMs, medical AI agents, and rater study design; and protein folding and design models and pipelines.
• Uncertainty estimation and factuality: conformal and Bayesian approaches to estimate uncertainty in deep learning and for LLMs, hypothesis testing, decision making under uncertainty, control variates, evaluation of natural language inference and LLM factuality.
• Computer vision: 3D vision, including generative models for 3D shapes and point clouds, and classical computer vision, including superpixel and image segmentation, keypoint and object detection/tracking.

On this webpage you will find articles, reading notes and most of my projects — which can also be found on GitHub or ShortScience . Here are some mission statements , as well as Twitter/X , LinkedIn , Xing , Google Scholar and YouTube profiles.

Previously, I completed my PhD at the Max Planck Institute for Informatics and both a bachelor and master degree at RWTH Aachen University with a short stint at Georgia Tech. Throughout this time, I also gained industry experience at Fyusion and Hyundai MOBIS and volunteered for the Max Planck PhdNet.

In a past career, I worked as a web engineer at Microsoft, RS Computer, and Fraunhofer FKIE using PHP, SQL and JavaScript to build web applications using Kohana and WordPress, among others.

I have been lucky to receive various awards for my work: At Google DeepMind, I have received patent awards in 2022, 2023, and 2024, as well as Google's Tech Impact award for SynthID in 2025.

For my PhD, I received the DAGM MVTec Dissertation Award 2023 and was supported by a Qualcommm Innovation Fellowship 2019. Furthermore, my work on bit error robustness was selected as outstanding paper at CVPR 2021 CV-AML and I was selected to participate in the 7th and 10th Heidelberg Laureate Forum 2019 and 2023 (with an Abbe Grant from the Carl Zeiss Foundation). I was also recognized as outstanding/top reviewer for CVPR, ICML and NeurIPS in several years.

For my MSc thesis, I received the STEM-Award IT 2018 as well as the Springorum-Denkmünze . My stint at Georgia Tech , was funded by the Hans Hermann Voss Foundation and my studies were supported by a Germany Scholarship. I have been featured on RWTH Aachen University's Dean's List for several years.