Feature stores have been traditionally designed for complex, Big-ML applications that normally assume that there is clear and high-ROI, advanced methods, and skilled staff, all resulting in long lead times. In this presentation, we cover Sub-ML – mid-complexity ML applications. In these, the uncertainty in terms of value, methods used, available staffing is higher, and speed is critical. We see Sub-ML rapidly growing across organizations and functions, which has led to a demand for a different feature store design that caters to the differences in the nature of the problems. In this presentation, we expand on our observations about the problem space, design constraints, and the thinking behind Enrich, our feature store for Sub-ML.