The increasing quantity of data from connected devices, sensors, and the Internet of Things (IoT) has made it necessary to move artificial intelligence (AI) processing from centralized cloud platforms to edge computing. Edge AI helps to run AI models on edge devices to solve the major issues of reliance on cloud-based AI, including high latency, limited bandwidth, and security issues. This paper gives a detailed survey of Edge AI, with focus on the advantages, architectural models, frameworks, and real-time applications in healthcare, autonomous vehicles, smart cities, and industrial IoT. In this study, a systematic approach was employed to compare Edge AI to the conventional cloud-based AI with the help of empirical analysis, case studies, and simulations. The results show that there is a significant improvement in the reaction time, bandwidth, energy consumption, and privacy. 
However, there are some challenges of Edge AI which include, limited resources, incompatibility issues, and security threats. The study looks at the new approaches such as federated learning, 5G-based edge computing, and decentralized AI to enhance the effectiveness and robustness of Edge AI. 
Future work suggestions include the optimization of existing AI models for low-power edge servers, integration of privacy-enhancing techniques, and the development of open and unified frameworks for efficient and secure Edge AI deployment. This work expands the current literature on Edge AI and discusses how it can improve the real-time decision-making processes across various industries.