The goal of this article is to develop an open unsupervised Machine Learning (ML) model that can classify groundwater sources as suitable or unsuitable for drinking, based on their hydrochemical parameters, without calculating their Water Quality Index (WQI) values, from Indian pumping stations. It evaluates its accuracy by comparing its performance with WQI models that are already established by previous researchers. It also has the capability to anticipate future contaminants by making the model adaptable and scalable. The work measures the WQI of 1083 groundwater stations in India using the Arithmetic Weighted method and verifies the unsupervised clustering model, the K-Means Clustering, in identifying the groundwater suitability of the pumping stations, without the knowledge of their WQI values. The WQI results indicate that most of the stations have excellent or good groundwater quality, while a few have poor or unfit groundwater quality. This knowledge also verifies the performance of the unsupervised learning model by validating its accuracy to 99.5% (for FIT) and 95.7% (for UNFIT) classifications. The group’s future work includes creating a colour-coded map of the groundwater quality with topographical maps and designing a Real-Time Monitoring system to prevent any harm from consuming unfit groundwater.