Air pollution is a major global health risk with significant impacts on human health and ecosystems. Among the key pollutants is fine particulate matter (PM2.5). Air quality monitoring is necessary to manage air pollution and evaluate the efficiency of pollution management strategies. Traditional ground-based monitoring is costly and offers limited spatial coverage, especially in low-resource areas, however, satellites provide global coverage of aerosol optical depth (AOD), which can be used to estimate PM2.5. This research aims to develop a robust machine-learning model to estimate PM2.5 levels from satellite data.