The availability of AI-powered drone systems has changed the game for autonomous painting and construction and maintenance works, delivering a breakthrough into efficient precision-based maintenance of such large structures. This work intends to develop an advanced architecture and a proper methodology for a drone system utilizing deep learning models in the autonomous execution of complex surface maintenance tasks. The requirements for safer, more accurate, and scalable industrial painting and cleaning solutions have led to increasing interest in AI-based drone technologies. The design of stable architectures for drones in unpredictable environments, along with their adaptability toward various surface materials, poses challenges for achieving high operational efficiency. This research work looks to improvise automation in surface maintenance with minimal human interventions and maximum accuracy along with safety. The architecture proposed draws on computer vision, sensor fusion, and path-planning algorithms for the identification of surfaces and their associated tasks. The dataset is downloaded for cracked and uncracked images and paint scrap the images were augmented and a dataset was made out of one image of paint. Testing revealed a high accuracy level in identifying surfaces and applying paints that achieved as high as 75% accuracy in a controlled environment. This research has transformative potential for autonomous maintenance using AI and may open up future applications in smart drones.