The construction industry significantly contributes to air pollution by releasing substantial quantities of pollutants that adversely affect both urban and natural environments. Among these pollutants, particulate matter (PM) emerges as a crucial concern. Such air pollution has broader negative consequences, impacting the health of workers and nearby communities, leading to environmental compliance costs, and threatening sustainability. Regulatory and research efforts have been initiated to tackle construction air pollution. However, the absence of comprehensive analysis of construction sites, ambient environmental data, and PM pollution hampers effective pollution detection and control measures. To address this challenge, this paper introduces an intelligent system that leverages machine learning and Internet of Things (IoT) technologies for monitoring and predicting PM concentrations associated with construction sites. The proposed system comprises three key components: (1) an IoT network for real-time data collection on ambient environmental conditions, construction-related air pollutants, and construction activities, (2) an objective detection model for detecting construction workers, vehicles, and heavy machinery that could contribute to air pollution, and (3) a deep learning-based predictive model to capture the relationships between air pollution and ambient environmental conditions as well as detected objects from construction sites. Experimental validation on a construction site at the Virginia Tech Blacksburg campus demonstrates promising results for the proposed system.