Prof. Dr. Jan Kieseler

Latest publications, talks and activities.

Jan Kieseler is an experimental particle physicist working at the interface of fundamental research and advanced technologies. His work focuses on top quark and Higgs boson interactions, detector optimization, and machine learning in reconstruction and computing. At KIT, he teaches courses in modern particle physics and data analysis, and supervises Bachelor, Master, and PhD students on topics ranging from collider physics to AI-driven detector design.

Contact:
Jan Kieseler (he/him)
Karlsruhe Institute of Technology (KIT)
Institute of Experimental Particle Physics (ETP)
Office: Building 30.23, Room 9-6
Office hours: by appointment only
Email: jan.kieseler@kit.edu

In the following, you can find the latest news, publications, talks and activities with collaborators and students.

April 23, 2025

End-to-End Detector Optimization with Diffusion Models

Kylian Schmidt, et al.: We introduce AIDO, a framework using diffusion models for optimizing particle detectors. A case study on sampling calorimeters shows how end-to-end optimization across simulation and reconstruction can guide detector design using learned surrogates.

March 18, 2025

End-to-End Optimal Detector Design with Mutual Information Surrogates

Kinga Wozniak, et al.: We introduce an AI-powered method that lets a computer optimise particle-detector layouts and quickly learn which designs capture the most useful information. Using a measure from information theory called mutual information, the approach finds detector configurations that match—or even beat—today’s best hand-tuned designs while requiring far less expert tweaking.

January 09, 2025

Constrained Optimization of Charged Particle Tracking with Multi-Agent Reinforcement Learning

Tobias Kortus, et al.: We introduce a multi-agent reinforcement learning approach with assignment constraints for reconstructing particle tracks in pixelated detectors. Our method collaboratively optimizes a policy to minimize particle scattering, employing a safety layer to ensure unique hit assignments and incorporating cost margins to enhance generalization. Empirical results on simulated proton imaging data demonstrate improved performance over existing single- and multi-agent baselines.

December 18, 2024

Toward the End-To-End Optimization of the SWGO Array Layout

Tommaso Dorigo, et al.: We present a continuous model of secondary particles from atmospheric showers and develop an optimization pipeline using stochastic gradient descent to identify the optimal layout for the SWGO water Cherenkov detector array. This approach aligns the detector configuration with the scientific goals of the experiment, demonstrating the capability to find global maxima in high-dimensional parameter spaces.