"The primary data in X-ray diffraction experiments are images which contain bright spots known as reflections. In order to compute the electron density of the crystal which yields these reflections, the intensity of each reflection must be estimated. Redundant observations must be parsimoniously reduced to a set of summary intensities and corresponding error estimates. This task is confounded by a range of systematic effects leading to equivalent reflections being expressed on different scales. "(Dalton et al. 2021) We would like to apply CARELESS as a deep learning method based on Bayesian statistics to merge and scale protein crystallographic data collected in sycrotrons and X-FELs. We have an ongoing collaboration with the creators of CARELESS.