Addressing real-world systems is often a multiscale endeavor where the macroscale aspects can only be fully appreciated with a clear understanding of the atomistic picture at their core. Using a combination of approaches appropriate for the relevant length and time scales, one can build a comprehensive description. An illustrative example is the reaction of water with zirconium alloys in nuclear reactors, causing oxidation, hydrogen pickup and embrittlement. Ab-initio calculations are employed to explore chemical reactions and to train a machine-learned potential (MLP) for the Zr-O-H system, which is used to compute materials properties feeding into a phase-field model. The MLP and phase-field models predict phase transitions in zirconia and successfully model the columnar growth of oxide films, as seen in experimental micrographs. All of the model building, computations, and analysis are carried out within the MedeA software environment, providing an efficient way to perform multiscale simulations.