At my company we have disjointed data stores for real-time ML inference vs. batch training/inference, causing user pain points when searching for features in different feature stores. I am excited to use Delta Lake and the Streaming Infrasturcture to make data more discoverable/usable for our ML use cases.