In this talk we will present our results using ab initio and machine learned models to discover candidate perovskite STWS materials. This includes predicting the stability of these materials at the operating temperature, whether these materials form as perovskites and in what specific phase, the energy of oxygen vacancies and other defects that mediate the redox process, and finally the viability of these metal oxides as STWS redox materials. The Gibbs energy, G, determines the equilibrium conditions of chemical reactions and materials stability. Unfortunately, G has been tabulated for only a small fraction of known inorganic compounds, thus impeding an analysis of whether materials are stable and synthesizable at operating conditions. We used a statistical learning approach to identify a simple and accurate descriptor to predict G(T) for stoichiometric inorganic compounds with ~40 meV/atom resolution, and minimal computational cost, for temperatures between 300 K and 1800 K. We then applied this descriptor to screen various compositions for their stability at STWS temperatures. Candidate materials that form as perovskites were then analyzed using a machine learned one-dimensional tolerance factor, τ, that we developed that correctly predicts 92% of compounds as perovskite or non-perovskite to predict which of the predicted stable compounds form perovskites. Finally, we applied DFT on the remaining viable candidates to predict the thermodynamics of STWS reactions. This requires predicting the correct perovskite phase, symmetry breaking distortions and modeling the correct defect, spin and charge state of the defect mediating the oxidation and reduction processes at the operating conditions.