Most machine learning (ML) platforms can readily change and test the model in the product. However, for some products, the model cannot be changed frequently or at all. For example, medical devices using ML models require rigorous review before the model can be changed. Similarly, models in self-driving cars and other robotics often have high-barriers to change. Companies that don’t yet have a product will also spend most of their time in a research setting. It’s important for these companies and institutions to have an ML platform with excellent research support. This presentation will describe how to create an ML platform that is research-centric. It will focus on how to make an ML research platform scalable, usable, and effective for supporting reproducible research.