The Savonius Hydro-Kinetic Turbine (SHKT) has a frugal design with the possibility of easy local manufacturing. Therefore, SHKT is a suitable proposition for off-grid power generation in standalone mode across the remote and hilly locations. In this work, an optimized geometry of a semicircular SHKT was proposed through 3D CFD based simulations, artificial neural network (ANN) augmented optimization and experiments. Firstly, CFD investigations of SHKT were performed to identify the parameters affecting the power coefficient (Cp). Results of CFD simulations were used to train ANN which was further used to optimize the blade parameters. Finally, experiments were conducted on the optimized blade to validate its performance. The results showed that aspect ratio between 1.4 and 2.0 and overlap ratio between 0.15 and 0.2 indicate better performance. Blade arc angle of 166° produced a maximum Cp of 0.194 at a TSR of 0.8. The study concluded that ANN is a time saving yet accurate tool for optimization of turbine blades, and the results provide a good agreement with the computational results with difference of 1.57% only. The optimized blade is found to be 8% more efficient than semicircular blades and is recommended for its applications in hydro farms and turbine clusters.