This research proposes a data-based, climate-resilient approach for identifying Groundwater Potential Zones (GPZs) in the semi-arid Nira River Basin (NRB) in the western region of Maharashtra (India). This study addresses the major challenge of unsustainable groundwater extraction exacerbated by hydro-climatic uncertainty and anthropogenic pressures. The present research examines long-term climatic trends for historic data (1950-2014) and projection data (2015-2100) based on Coupled Model Intercomparison Project Phase 6 (CMIP6) projections according to the Shared Socio-economic Pathways 2 (SSP2) 4.5 scenario. It reports an increase in temperature coupled with minor changes in rainfall, resulting in increased aridity over the NRB. Further, the present research applied an ensemble machine learning (ML) modeling strategy that combines Artificial Neural Networks (ANN), eXtreme Gradient Boosting (XGBoost), and Light Gradient Boosting Machine (Light GBM). The XGBoost model showed better predictive performance (AUC = 0.97), identifying rainfall as the most dominant factor, followed by elevation and land surface temperature (LST). The models' robustness was validated based on observed groundwater well data (2023). Prioritization of the most important factors for the identification of water-stressed regions (412.65 km2) was executed using the Analytical Hierarchy Process (AHP). The Consistency Ratio (CR) was determined to be 0.09, indicating an acceptable consistency level and enabling strategic decision support for long-term water security. The 81 precise locations were determined through the spatial intersection in the waterstressed areas, where the groundwater management interventions would be applied. These sites serve as the foundation for a scalable framework aimed at promoting sustainable groundwater management in study regions.