Model management is a piece of the data science life cycle that helps organizations consistently and safely develop, validate, deliver, and monitor models to create a competitive advantage. Previously, model management referred to the monitoring of production models. Dell EMC and Domino believe that model management must encompass a much broader capability.
Models involve code and data, but they are fundamentally different from software or data. A common misconception is that because models involve code and data, you can treat them as if you are developing a software application or storing data for archival purposes. This misconception prevents data scientists from unlocking the full potential of models. Models involve a learning algorithm and its associated weights, which are updated as the model is trained. Existing tools such as Git, Jenkins, and artifact repositories are not designed to capture the context and governance that is essential in managing a model through its life cycle.
Model management has five key pillars. Each is necessary but insufficient in isolation to unlock the full power of models:
Domino encompasses all these model management pillars to streamline the model management process to save customers time and resources.