Demonstration of MLflow : A System to Accelerate the Machine Learning Lifecycle

Machine learning development creates new challenges that are not present in a traditional software development lifecycle. These include keeping track of the myriad inputs to an ML application (e.g., data versions, code and hyperparameters), reproducing results, and production deployment. Existing systems built to address these challenges restrict the supported programming languages and ML libraries that developers can use to train and deploy models, creating difficulties for organizations that must leverage a variety of ML libraries and model deployment environments. Accordingly, we propose to demonstrate MLflow: a system that streamlines the machine learning lifecycle and is designed to work with any ML library, algorithm, or programming language. Available at mlflow.org, MLflow is an open source project with over 70 contributors. Through our demonstration, audience members will interact with each of MLflow’s major components, experiencing firsthand how the platform’s open interface enables developers across an organization or research group to collaborate on reproducible machine learning workflows that leverage their preferred languages and ML libraries.