Edge devices, which include a wide range of hardware from smartphones to IoT sensors, wearables, and autonomous vehicles, exhibit considerable heterogeneity in terms of computational power, memory, network capabilities, and data availability. This diversity introduces several challenges for distributed machine learning frameworks, such as federated learning. In this paper, we integrated the actor model of concurrency with federated learning, and supported the distributed learning process using asynchronous communication. Our experimental results show this approach has great potential of achieving better performance when executing on heterogeneous edge devices. Using the actor model as the underlying computational platform, we expect that decentralized machine learning paradigms can become more efficient and widely applicable across diverse and large-scale edge AI applications. This shift toward asynchronous and distributed learning architectures is crucial for the future of personalized AI, industrial automation, healthcare, and autonomous systems, where real-time learning and adaptability are essential.