Project partners
Karlsruhe Institute of Technology, Institute for Experimental Particle Physics (ETP)
Prof. Dr. Torben Ferber (project coordination)
Wolfgang Gaede Strasse 1
76131 Karlsruhe, Germany
Email: torben.ferber@kit.edu
ETP develops algorithms for detector reconstruction and calibration in collider experiments with a strong focus on software quality, performance, and reliability. The group leads calorimeter software and AI-based track reconstruction at Belle II.
Work spans classic reconstruction methods and modern AI driven techniques for event interpretation.
The team demonstrated the first full AI based tracking solution for a large drift chamber and operates the first GNN-based calorimeter trigger in the Belle II hardware trigger system since 2025.
The group combines physics insight with practical deployment at scale, including leadership roles in trigger and software coordination at Belle II.
Karlsruhe Institute of Technology, Institute for Information Processing Technology (ITIV)
Prof. Dr. Ing. Dr. hc. Juergen Becker
Engesserstrasse 5
76131 Karlsruhe, Germany
Email: becker@kit.edu
ITIV focuses on reconfigurable hardware and embedded systems with expertise in architecture design, methods and tools, and hardware and software co design. The group builds domain specific architectures and end to end tool flows for efficient AI on heterogeneous platforms.
Activities range from high rate data acquisition and control to energy aware acceleration on FPGA and SoC devices.
Many years of collaboration with ETP targets scalable deployment of dynamic Graph Neural Networks on modern adaptive devices with practical attention to partitioning and compiler support.
University of Bonn, Research and Technology Center for Detector Physics (FTD)
Prof. Dr. Klaus Desch
Kreuzbergweg 24
53115 Bonn, Germany
Email: desch@uni-bonn.de
The Bonn teams develop advanced detector systems and real time reconstruction methods with strong contributions to tracking and data acquisition. Work includes firmware and FPGA based readout, trigger design, and reconstruction algorithms for modern spectrometers.
Recent efforts evaluate adaptive devices and Graph Neural Network inference for tracking on new hardware.
The group links detector development with realistic online processing constraints to improve selection quality under high data rates.
University of Freiburg, Department of Physics
Prof. Dr. Marco Gersabeck
Hermann Herder Strasse 3
79104 Freiburg im Breisgau, Germany
Email: marco.gersabeck@physik.uni-freiburg.de
The Freiburg group brings long standing expertise in real time analysis, detector alignment, and high rate data processing at LHCb. The team achieved micron level alignment of the VELO and helped establish real time calibration that enabled analysis quality data directly from the online system.
They contributed to GPU based event selection and to FPGA based concepts that reduce load on later stages.
The group bridges algorithms and heterogeneous computing with an emphasis on robust operation at collider scale.
Technische Universitaet Hamburg, Institute of Control Systems
Prof. Dr. Ing. Annika Eichler
Harburger Schlossstrasse 22a
21079 Hamburg, Germany
Email: annika.eichler@tuhh.de
The TU Hamburg group works on data driven control and diagnosis for complex systems with strong links to DESY. The team studies reinforcement learning and optimization for accelerator tuning and develops detection and classification methods for machine protection.
Research connects physics models with machine learning and practical firmware integration.
Close collaboration with accelerator control groups enables methods that are both effective and deployable in real operations.
SiMa Technologies Germany GmbH
Dr. Ing. Karim Guirguis
Friedrichstrasse 15
70174 Stuttgart, Germany
Email: karim.guirguis@sima.ai
SiMa.ai develops embedded machine learning platforms for edge AI with a software first approach. The MLSoC and Modalix product families aim for high performance and energy efficiency across diverse sensors and models.
The company supports smooth deployment from development to production on constrained devices.
Collaboration focuses on practical toolchains and runtime support that meet latency and throughput targets in real applications.
