Decentralized and edge AI systems aim to shift computation from centralized data centers to local devices such as smartphones, IoT sensors, and autonomous vehicles. This architecture reduces latency, enhances privacy, and minimizes dependency on continuous internet connectivity. It relies on lightweight optimized models capable of operating under constrained hardware conditions. A key component of this approach is federated learning, where multiple devices collaboratively train a shared model without exchanging raw data. This significantly improves data privacy and system scalability. This paradigm is becoming essential for real-world AI applications such as smart cities, autonomous transportation, and advanced healthcare systems.