Typical model validation exercises in fusion research applications are characterized by computationally costly forward models with a selection of uncertain, phenomenological input parameters that are constrained only implicitly through their impact on the model predicted measured quantities. Often the dimensionality of this optimization task is large enough that the approach of manually calibrating these free input parameters is untraceable. Bayesian inference and optimization algorithms could potentially provide a principled and data-efficient approach to constrain these free parameters without human in the loop, such that the scientist can focus on the creative science analysis task. In this work, the Bayesian approach is applied for DREAM simulations of runaway electron generation in JET after an argon Massive Gas Injection [1, 2]. The present focus of the project is to apply this methodology to constrain the uncertain Re seed distribution in the simulations such that the resulting synthetic synchrotron radiation distribution, calculated with SOFT [3], matches the experimentally measured synchrotron radiation emission distributions. The distance between measured and synthetic images is quantified by first realigning the two images, such that the peak emission occurs near the center of the image, to compensate any potential misalignment due to, e.g., magnetic reconstruction inaccuracies during the disruption, and then calculating the maximum mean discrepancy [4] between the two images. Since the kinetic DREAM simulations with SOFT post processing are computationally quite expensive, requiring several CPU hours for each sample, batch Bayesian optimization is key to reduce the overall wall clock time for the optimization task to acceptable values [5, and references therein]. By applying this methodology, it will be assessed to what extent the model is able to capture the experimentally measured time-evolution of the synchrotron emission distribution during the early part of the RE plateau with a given initial RE seed distribution. [1] A.E. JÄRVINEN, et al. J. Plasma Phys. 88 (2022) 905880612. [2] M. HOPPE, et al. Comput. Phys. Commun. 268 (2021) 108098. [3] M. HOPPE, et al. Nucl. Fusion 58 (2018) 026032. [4] A. GRETTON, et al. J. Machine Learning Res. 13 (2012) 723-773. [5] N. Hunt, 2020 ‘Batch Bayesian Optimization’ Master’s thesis, Massachusetts Institute of Technology.