The Helmholtz AI Conference (HAICON26) brought together more than 600 researchers, industry partners and representatives from 21 countries in Munich in early June. Under the theme “AI for Science”, participants discussed current developments at the intersection of artificial intelligence and scientific research. SCC contributed numerous activities and shared its expertise in research infrastructures, MLOps, federated learning, quantum machine learning, and AI applications for energy, climate and safety research.
Already on the conference's Workshop & Tutorials Day, SCC was represented with an invited talk on the international benchmarking initiative UNLOCK. The presentation highlighted experiences and challenges in evaluating modern AI systems. The contribution emerged from the work of the Junior Research Group Robust and Efficient AI (RAI, headed by Charlotte Debus) and was presented by Philipp Huber.
Another focus area was quantum machine learning. Two tutorials introduced participants both to the theoretical foundations and to practical programming approaches for applying quantum-based methods in machine learning. The sessions were organised by Eileen Kühn and Gabriel Mejia Ruiz from SCC's Quantum Machine Learning team.
Questions surrounding the productive development and deployment of AI were addressed in two World Café sessions. Participants discussed federated learning, which enables the collaborative development and fine-tuning of large AI models across institutional boundaries without the need to share data. A second World Café focused on tools and requirements for effectively managing the entire AI and machine learning lifecycle. These sessions were primarily organised and moderated by Khadijeh Alibabaei, Leonhard Duda, Lisana Berberi, Valentin Kozlov and Paul Kieckhefen.
The SCC poster presentations, presented by Lisana Berberi, Khadijeh Alibabaei and Paul Kieckhefen, reflected the broad spectrum of SCC's research and infrastructure activities. Topics included a pilot MLflow service for Helmholtz researchers, the recently launched European EOSC-ARENA initiative for building an open and ethically responsible generative AI infrastructure for research, and new surrogate-model approaches for the safety analysis of severe nuclear accidents.
Additional contributions by Elena Vollmer, Muhammed Öz and Kaleb Phipps focused on data-driven applications in energy and environmental sciences. The presented work covered RenewBench, a dataset for energy systems research, differentiable power-flow optimisation for electricity grids, and innovative AI-assisted approaches for interacting with global weather models.
The contributions of Lisana Berberi and Elena Vollmer were additionally selected for Poster Spotlight Talks, highlighting their particular relevance and impact.
Through this broad engagement within the Helmholtz AI community, SCC researchers actively contributed to shaping interdisciplinary exchange around “AI for Science” at HAICON26.