Towards automated modeling of group and cluster scale strong lenses in LSST

The Legacy Survey of Space and Time (LSST) will vastly increase the number of known stronglenses, demanding scalable and automated modeling solutions. While most existing efforts focus ongalaxy-scale lenses, the LSST will reveal many group and cluster-scale lenses, which are morecomplex in their mass distribution, requiring novel modeling methods. This study aims to develop andvalidate an algorithm for the efficient modeling group and cluster-scale strong lens systems featuringextended arcs using ground-based data. Our algorithm builds on existing software for galaxy-scalelens modeling, incorporating halo and subhalo components to account for complex mass distributions.To enhance computational efficiency we employ GPU parallelization, an inference method thatcombines image position and pixelated data, and a Taylor expansion for the subhalo massdistribution. We validate our algorithm using both simulated and real lenses. Our method successfullymodels complex lens systems within a minute, accurately recovering lens parameters and theirassociated uncertainties.