Whenever you're ready, walk me through how you'd approach this.
How would your feature engineering strategy change if this data was streaming in real-time instead of arriving in a static batch?
