Feature Store Summit - Feature Store at Varo: Why, How and Lessons Learned

This presentation is part of the Feature Store Summit.
----
Putting ML models in production requires many engineering components to work in perfect sync and feature stores are one of the key components of this complicated machinery. Feature stores are responsible for providing necessary input data both at training and in production. When it comes to designing and developing feature stores, however, there is no single approach. The best path forward depends on the organization, technology stack, resources, use-cases, and several other factors. In this presentation, we walk through the evolution of the feature store at Varo starting from the ontological challenges of defining the feature store and the key functionality that it needs to support.