Event
On June 4th, 2025, we had the pleasure of joining this year’s Helmholtz AI Conference in Karlsruhe — a key gathering for anyone working at the intersection of artificial intelligence and scientific research. Representing Perelyn was our colleague Michael Banf, who took part in Poster Session II and shared some of our latest research with the community.
Michael presented our poster, “A Topological Remedy for Over-Squashing in Graph Learning via Forman‐Ricci Curvature based Graph-to-Hypergraph Structural Lifting.” This work introduces a new approach to tackling one of the major challenges in graph neural networks: maintaining long-range information without it getting compressed or lost — a problem known as over-squashing. By using ideas from differential geometry and topology, the method shows promise for improving how GNNs handle complex systems, from biology to physics and beyond.
It was a fantastic opportunity to meet and exchange ideas with researchers tackling similar problems across a wide range of disciplines. The conversations we had — both during the session and throughout the event — were thoughtful, energizing, and genuinely inspiring.
Thanks to everyone who stopped by for a chat, shared insights, or just came to learn more. We’re excited about the potential collaborations ahead and look forward to staying in touch with the Helmholtz AI community.
The poster presented at the conference is shown below.
News
The proceedings from the "AI in Production" workshop at KI2025 have been published. Among the contributions is a paper by our team on the use of the Model Context Protocol (MCP) in industrial production environments.
News
Dominik Filipiak and Michael Banf are co-authors of a community paper on the 2025 Topological Deep Learning Challenge, published in the Proceedings of Machine Learning Research. Their contributions feed into TopoBench, an open benchmarking library for the research community.
