This study explores the use of Artificial Intelligence (AI) and edge computing to improve the performance of cloud-native AI applications. Traditional cloud-based AI suffers from latency and a reliance on centralized resources, which hinders its viability and effectiveness in real-time systems and resource-constrained systems, including Internet of Things (IoT) systems, autonomous vehicles, and smart cities. The major problem being addressed is that traditional cloud infrastructure cannot provide the low-latency and high-efficiency necessary to run modern AI applications. In this research, the deployment of smaller AI models (MobileNet, Tiny-YOLO, decision trees, and federated learning) and their performance accuracy, response time, and resource utilization on edge devices were explored. With the IoT-2 Dataset and a simulated edge environment, this study shows that edge-based AI execution can reduce latency by 25% and increase resource efficiency by 15%, compared to cloud-based counterparts or systems used for comparisons. MobileNet had the highest accuracy (92%), and Tiny-YOLO had the fastest response time (120 ms), essential for latency-sensitive tasks. The study showed better privacy and minimal resource consumption for federated learning. This study indicates that AI-powered edge computing can help scale, increase real-time decision-making capacity, and markedly reduce reliance on cloud infrastructures in distributed systems.