Building a Modular AI Deployment Framework for Model Sharing Across Cloud Environments

Abstract

The proliferation of AI models across industries has spurred the need for flexible and interoperable deployment strategies that enable seamless migration and sharing across heterogeneous cloud environments. This paper proposes a modular AI deployment framework that decouples model development, packaging, and orchestration layers to ensure cloud-agnostic portability. Leveraging containerization, API standardization, and automated orchestration tools like Kubernetes, the framework supports scalable deployment with enhanced reproducibility and reduced vendor lock-in. Evaluation across AWS, Azure, and GCP environments demonstrates the framework's efficiency, adaptability, and cost-effectiveness.