In response to regulatory requirements and to make their data science team more productive, PayPal developed Quokka, a platform that implements a CI/CD and shadow deployment process for ML models and model features. This platform was designed to streamline model testing and deployment workflows and prevent problems that might arise due to distribution shift or runtime system issues. Quinn Zuo reviews the platform architecture, which includes UIs for scheduling deployment, an automated CI/CD pipeline, and an auditing compute environment for shadow testing. She describes how audited models may graduate from shadow to production once certain metrics are collected and evaluated.