Speakers
Description
Blop is a BeamLine OPtimization package built on top of BoTorch, a Bayesian Optimization framework, with Bluesky in mind [1]. It proved to be very useful to align and tune synchrotron and electron beamlines at the NSLS-II and ATF user facilities of the Brookhaven National Laboratory, ALS of the Lawrence Berkeley National Laboratory, and LCLS of the SLAC National Accelerator Laboratory. The tool is particularly useful for the beamlines which already have the Bluesky/Ophyd setup – the integration of Blop there is very straightforward.
Recent developments in Blop are making significant progress. The default backend framework for Bayesian optimization will soon be based upon Meta’s Ax platform [2]. This provides an adaptive experimentation framework with many built-in features for deployment, visualization, performance tracking, logging, and automatic model selection. In addition, Blop will soon have standard support for Tiled, Ophyd-Async, and Bluesky Adaptive built-in. Future developments will focus on solving complex beamline optimization problems through simulation via XRT, making safe Bluesky optimization routines, and enabling collaborative optimization between users and Blop agents.
References
[1] T. W. Morris, M. Rakitin, et al. “A general Bayesian algorithm for the autonomous alignment of beamlines,”
Journal of Synchrotron Radiation, vol. 31, no. 6, pp. 1446–1456, Nov 2024.
https://doi.org/10.1107/S1600577524008993
[2] M. Olson, E. Santorella, et al. “Ax: A Platform for Adaptive Experimentation,” AutoML 2025 ABCD Track,
2025. https://ax.dev/
