This work explores the feasibility of modernizing the current leak detection method used by the City of Kitchener which involves manual acoustic readings performed on a third of the city annually. We seek to develop software to detect leaks in real time using pressure and flowrate data collected by remote sensors in water pipelines. The primary objective is to update the detection to be in real time and increase sensitivity in the process by detecting smaller leaks that could have previously gone undetected. We have decided to achieve this using a time-series classification algorithm: MLSTM-FCN and the LeakDB dataset to represent a scaled-down version of the water distribution network in the City of Kitchener. The configuration of using pressure sensors only was selected from the results of the reduced feature test. It provided satisfactory performances in the proceeding generalization and localization tests. The solution fulfills all constraints and criteria. Based on the analysis, it is recommended to install 388 pressure sensors in the City of Kitchener as it minimizes the cost, without sacrificing the accuracy of the model.