ABSTRACT Hydrokinetic turbines can generate electricity in remote rural locations using the kinetic energy of nearby rivers or canals. The Savonius hydrokinetic turbine (SHKT) is the easiest to design and manufacture. Optimization of blade shape factor (p/q) and blade arc angle (Φ) can contribute significantly to enhancing the efficiency of a turbine. The present work proposes a time-saving and reliable method to design an optimized SHKT by using a blend of experimentally validated 3D computational fluid dynamics (CFD) simulations and Artificial Neural Network (ANN). To optimize the turbine blade, the turbine performance needs to be analyzed for a number of values of blade parameters selected at very small intervals. Performing so many CFD simulations is a costly task. Application of an ANN tool trained using a smaller number of CFD results should significantly curtail the costs while maintaining reliability as well. The power coefficient (Cp) of SHKT was obtained using CFD simulations for some selected sets of Φ and p/q. These results were used to train the ANN by creating a parametric map between the input parameters viz. Φ, p/q, and the output parameter Cp. The trained ANN tool was further used to predict the turbine’s performance for sets of input parameters varying at very small intervals. The blade with p/q of 0.2 and Φ of 148° provides a maximum Cp of 0.209 at a TSR of 0.8. This optimal blade was 10.5% more efficient than the standard semicircular blade.