Controlling and monitoring groundwater quality is a crucial and often complex task to ensure there is no contamination in. To this end, we studied and modeled the transfer of contaminants in the Triffa Plain, as a case study. We used a crafted dataset with different features (such as well depth, location (longitude and latitude), season (spring, summer, autumn, winter), temperature, precipitation, crop type, nitrate concentration, pesticide concentration), gathering data from several Triffa’s wells. This dataset served as the guide for supervised learning, allowing us to train AI models designed to predict the contamination rates at any point on the plain. Indeed, we trained and tested four regression models: Random Forest, XGBoost, LightGBM, and CatBoost. Based on the results obtained using the Root Mean Square Error (RMSE) metric, we found that Random Forest outperformed the other models. This best-performing model was then deployed online on a real interactive map of the Triffa Plain. By clicking on any point on the map or entering its coordinates (longitude and latitude), the predicted contamination rate (the concentration of nitrates (NO₃⁻) and pesticides (atrazine or others)) are displayed in green or red depending on the severity of contamination.