Scientific AI
We develop principled AI methods to solve hard problems from the natural sciences. Our expertise are fundamental algorithms and their application to complex problems with spatial structure.
02.2026 Our recent function-centric graph neural network establishes a new state of the art for predicting the electronic ground state density of molecular systems! See Manuel & co. upcoming ICLR 2026 paper openreview.net
12.2025 Simon Wagner has been selected as one of the Top Reviewers at NeurIPS 2025 😊
Highlights
Geometric Machine Learning in Quantum Chemistry
Predicting molecular properties is important for a huge range of problems, from biochemistry to drug development.
The most universal approach to such predictions relies on quantum mechanics. We aim to dramatically speed up these quantum calculations using machine learning methods that respect the fundamental symmetries of the problem.
Our Team
The team is everything! Our maxime: 1. Be Nice and 2. Do Great Science :-)
Our Network
We learn a lot from our colleagues! We are proud members of, and help shape the future of, our excellence cluster, our ellis unit and our department.
Home, Sweet Home
IWR is a central institute of our university, and a fantastic place where scientists from maths, CS, physics and more meet to develop methods that address some of the hardest problems out there.
Machine Learning in Heidelberg
Heidelberg is home to a vibrant ML community. The MLAI page is the portal to all talks and events in the area.