Bernhard’s research spans autonomous driving, reinforcement learning, and machine learning. He recently became part of the founding team of KE:SAI, a new Franco-German research initiative dedicated to advancing open science in physical AI. His involvement reflects both his scientific contributions and his role in helping build one of Europe’s newest AI research organizations.
We are thrilled to feature Bernhard Jaeger as our next IMPRS-IS Scholar Spotlight! Bernhard joined IMPRS-IS in 2022 as a doctoral researcher in Andreas Geiger’s Autonomous Vision Group in Tübingen. During his doctoral studies, he has conducted research at the intersection of autonomous driving, reinforcement learning, and machine learning, while also contributing to one of Europe’s newest AI research initiatives.
In 2026, Bernhard became part of the founding technical team of KE:SAI (Kyutai ELLIS Scalable Autonomous Intelligence), a newly launched non-profit research laboratory based in Tübingen and Paris. The initiative is a collaboration between Kyutai and the ELLIS Institute Tübingen and is dedicated to advancing open science in physical AI, intelligent systems that can understand and act in real-world environments. Together with IMPRS-IS faculty Andreas Geiger and Bernhard Schölkopf, as well as fellow IMPRS-IS scholars Kashyap Chitta and Daniel Dauner, Bernhard is helping establish a research organization focused on autonomous driving, physical AI, and causal world models.
KE:SAI aims to develop a fully open technology stack for embodied intelligence, combining scalable simulation, world models, and robotic learning systems. One of its first major goals is the development of a fully open self-driving stack, while its longer-term vision extends to broader robotics applications in domains such as manufacturing and healthcare. Bernhard’s involvement highlights the growing role of doctoral researchers not only in advancing scientific research but also in helping build new institutions that shape the future direction of AI.
Alongside these activities, Bernhard has built an impressive research portfolio during his doctorate. One of his primary research directions has been understanding the limitations and hidden failure modes of end-to-end driving systems. His first-author ICCV 2023 paper, Hidden Biases of End-to-End Driving Models, investigated systematic biases that can emerge in modern autonomous driving systems and highlighted challenges that must be addressed before such models can be safely deployed at scale.
Beyond autonomous driving, Bernhard has also made contributions to reinforcement learning. His publication, An Invitation to Deep Reinforcement Learning (Foundations and Trends in Optimization, 2024), provides a comprehensive overview of the field, while his CoRL 2025 paper CaRL: Learning Scalable Planning Policies with Simple Rewards introduced a framework for learning planning policies that scale effectively while relying on simple reward signals.
In addition to his first-author work, Bernhard has collaborated on several influential projects, including TransFuser: Imitation with Transformer-Based Sensor Fusion for Autonomous Driving (IEEE TPAMI 2023), GTA: A Geometry-Aware Attention Mechanism for Multi-View Transformers (ICLR 2024), End-to-End Autonomous Driving: Challenges and Frontiers (IEEE TPAMI 2024), and LEAD: Minimizing Learner-Expert Asymmetry in End-to-End Driving (CVPR 2026). In particular, both TransFuser and the End-to-End Autonomous Driving survey have become widely referenced within the autonomous driving community.
Before joining IMPRS-IS, Bernhard earned a B.Sc. in Informatics: Games Engineering from the Technical University of Munich. He subsequently completed an M.Sc. in Computer Science at the University of Tübingen, where he further developed his interests in machine learning and artificial intelligence.
His achievements have also been recognized through competitive funding. In 2025, Bernhard received a grant of over €90,000 from the Vector Foundation to support a research project on reinforcement learning for autonomous driving. The award enables him to explore new approaches for applying learning-based decision-making methods to autonomous systems and further advance research at the intersection of artificial intelligence and autonomous driving.
Congratulations, Bernhard! We are proud to feature your achievements in this IMPRS-IS Scholar Spotlight!
Co-founders of KE:SAI. From left to right: Kashyap Chitta, Daniel Dauner, Andreas Geiger, Bernhard Schölkopf, Bernhard Jaeger
More information