Improving the Reliability of Cloud-Based Pre-Trained Machine Learning Models

This thesis identified non-trivial implications to software quality resulting from evolving pre-trained machine learning models served through cloud computing platforms. The study's technological solutions help resolve challenges faced by developers dependent on such 'plug-and-play' technology, both through a novel software architecture that enhances integration robustness and improved service documentation techniques.