Maximizing the Power of Cognitive Services with an AI-as-a-Service Architecture for Seamless Delivery

The empowerment of edge devices with intelligence presents both challenges and opportunities, primarily due to their computational capacity while maximizing user satisfaction, which necessitates seamless approaches. Currently, various state-of-the-art approaches and standard bodies aim to standardize and develop solutions that enable intelligent edge computing from different perspectives. Despite these efforts, leveraging networks that are aware of delivering cognitive services to end users remains a challenge, particularly in embedded devices. This study presents an innovative approach, Artificial Intelligence as a Service (AIaaS), to effectively manage the lifecycle of diverse AI models and paradigms while offering them as a service for heterogeneous devices. Among our findings, we validate the suitability of our AIaaS Architecture for seamlessly handling training jobs in a federated learning case study on heterogeneous devices.