Karsten Roth
Research Scientist at Google DeepMind
Tübingen, Baden-Württemberg, Germany
Summary
👤
Senior🎓
Top SchoolKarsten Roth is a research scientist at Google DeepMind with nine years of experience bridging foundational multimodal models and post-training methods. He completed a summa cum laude PhD through IMPRS-IS and ELLIS under Zeynep Akata and Oriol Vinyals, focusing on continual learning, OOD generalization, contrastive learning and multimodal pretraining. His background spans top research labs—including Meta AI, Amazon (where he helped develop PatchCore for industrial anomaly detection), Mila and the Vector Institute—reflecting a strong track record in representation learning for both medical and industrial applications. Trained originally as a physicist in Heidelberg, he blends rigorous analytical thinking with practical impact in large-scale model development and transfer learning. Colleagues describe him as someone who consistently turns theoretical insights into robust, deployable techniques for model generalization.
9 years of coding experience
4 years of employment as a software developer
Doctor of Philosophy - PhD, Summa cum laude, Doctor of Philosophy - PhD, Summa cum laude at University of TuebingenMaster's degree, Physics, 1.1 (Best 1.0), Master's degree, Physics, 1.1 (Best 1.0) at Heidelberg University
German, English, Chinese, French
Github Skills (38)
meta-learning10
metric-learning10
image-retrieval10
wandb10
generalization10
recall10
arxiv10
evaluation9
deep-learning8
keras8
der8
pytorch8
reinforcement-learning7
api7
hyperparameter-tuning7
Programming languages (2)
Github contributions (5)
Confusezius/Revisiting_Deep_Metric_Learning_PyTorch
May 2020 - Feb 2021
(ICML 2020) This repo contains code for our paper "Revisiting Training Strategies and Generalization Performance in Deep Metric Learning" (https://arxiv.org/abs/2002.08473) to facilitate consistent research in the field of Deep Metric Learning.
Contributions:34 commits, 30 pushes, 1 branch in 9 months
meta-learningmetric-learningtrainingicmlarxiv
Confusezius/Simulated_SelfDriving_Car
Jan 2019 - Jan 2019
Contributions:1 commit, 1 branch in 1 day
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.