Session: What’s New with rubicon-ml? Capital One’s Solution for Logging the ML Lifecycle
In this talk sponsored by Capital One’s Open Source Program Office, participants will learn more about rubicon-ml, a machine learning solution developed to standardize the model development lifecycle. Based on Python, rubicon-ml leverages popular open source libraries like fsspec, intake, and dash to enable easy logging, visualization, and sharing of experiment metadata. Users can range from machine learning practitioners, collaborators and reviewers as they train, monitor and govern models that solve complex business problems.
This session will show users how to incorporate rubicon-ml into their existing machine learning workflows. Users will see how rubicon-ml plays nicely with other commonly used open source Python libraries like Scikit-learn to train models and Dask to distribute that training. We will also show how rubicon-ml schema can be used to standardize metadata logging of open source models from packages like Scikit-learn, XGBoost, and LightGBM among others.
The intended audience for this talk is machine learning practitioners of any level – from developers, to data scientists, to model review and risk officers. rubicon-ml is designed to help multiple user profiles track model iteration across all stages of the machine learning lifecycle.
