Flux: Groupon's automated, scalable, extensible machine learning platform

As machine learning becomes the driving force of the daily operation of companies within the information technology sector, infrastructure that enables automated, scalable machine learning is a core component of the systems of many large companies. Various systems and products are being built, offered, and open sourced. As an e-commerce company, numerous aspects of Groupon's business is driven by machine learning. To solve the scalability issue and provide a seamless collaboration between data scientists and engineers, we built Flux, a system that expedites the deployment, execution, and monitoring of machine learning models. Flux focuses on enabling data scientists to build model prototypes with languages and tools they are most proficient in, and integrating the models into the enterprise production system. It manages the life cycle of deployed models, and executes them in distributed batch mode, or exposes them as micro-services for real-time use cases. Its design focuses on automation and easy management, scalability, and extensibility. Flux is the central system for supervised machine learning tasks at Groupon and has been supporting multiple teams across the company.