<p>Optimizing cellular networks is essential for improving coverage and service quality, with a particular focus on the orientation of sector antennas at radio sites, i.e., azimuths. Although this parameter is crucial, its management is complex. Engineers often rely on default configurations and adjust azimuths based on drive test results, a method that can be time-consuming and costly, requiring frequent manual adjustments. To overcome these challenges, we propose a hybrid method combining AI and EM wave propagation models within a GIS. This method consists of two modules: an AI module using a genetic algorithm to generate solutions based on past performance, and a GIS module to evaluate these solutions using EM propagation models. While these modules operate independently, they exchange data via intermediate files. This approach reduces both the time and cost of optimization while improving radio coverage. Applied to 4 LTE sites of the ATM Mobilis operator in a study area covered by 11 radio sites, this method improved network coverage by 3.13% compared to parameters previously set by the radio team and by 6.09% compared to default configurations. These results clearly highlight the effectiveness and potential of this approach.</p>