CAREER: ElasticCML: Elastic Framework for Collaborative Machine Learning in Multi-Cloud Environments

This project addresses the growing challenges in collaborative machine learning (CML) across industries like healthcare and finance, where participants jointly develop machine learning (ML) models while preserving data privacy. Current CML systems struggle with participant heterogeneity, varying network conditions, and diverse pricing models in multi-cloud environments. To tackle these challenges, we present ElasticCML, an integrated framework that introduces three key innovations: intelligent resource management, efficient distributed training strategies, and dynamic cloud resource orchestration. ElasticCML features adaptive mechanisms that automatically optimize resource allocation based on participants' contributions and capabilities. Its communication layer intelligently reduces data transfer overhead while preserving model quality. The framework also implements smart scheduling algorithms that minimize operational costs across multiple cloud platforms. These components work together to optimize resource utilization while maintaining model training performance under diverse infrastructure conditions. The framework continuously adapts to changing system dynamics and varying computational capabilities of participants, ensuring efficient and inclusive ML development. The project's broader impact lies in democratizing advanced ML capabilities, particularly in sensitive and resource-constrained domains. Through open-source contributions, educational programs, and workshops, ElasticCML aims to advance resource-efficient AI development while fostering innovation across academia, industry, and society. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.