Artificial intelligence for gravitational-wave experiments

Wednesday 18 Nov 2026, 11:00 → 12:00 Europe/Rome

Gravitational-wave detectors are the most finely controlled machines ever built: hundreds of coupled control loops keep them running at unprecedented sensitivity. Artificial intelligence is now being used across detectors and their R&D facilities to characterize noise, clean data, build virtual sensors, control the interferometer in real time, assist its operation, and even search for new detector designs. I will give an overview of this work. I will then turn to my own work on reinforcement-learning control. Deep Loop Shaping, designed in a collaboration between GSSI, Caltech, and Google DeepMind, reduced control noise at LIGO Livingston by more than a factor of 30, and up to a factor of 100 in parts of the 10–30 Hz band, where better sensitivity benefits intermediate-mass black holes and early warning of neutron-star mergers (Science, 2025). I will describe its first shadow-mode tests on Virgo and the path to full deployment, our plans to test ML-based control at the GEMINI underground testbed, and what machine learning could mean for the Einstein Telescope. I will close with the current activities of the AI-for-ET division.

Tomislav Andric is a fixed-term researcher at the Gran Sasso Science Institute in L'Aquila. His work is on the control of gravitational-wave interferometers, in particular reinforcement learning for detector control. He contributed to the first demonstrations of reinforcement-learning control in gravitational-wave interferometers at the Caltech 40 m prototype, GEO600, and LIGO Livingston (Science, 2025). For the Einstein Telescope, he works on seismic isolation, inter-platform control, and Newtonian noise. He obtained his PhD at GSSI in 2023 and was a postdoc at the Max Planck Institute for Gravitational Physics in Hannover. He co-chairs the Inter-Platform Motion Working Package of the ET Instrument Science Board and the LVK Machine Learning Algorithms Working Group, and co-coordinates the AI-for-ET Division of the Einstein Telescope Collaboration.