How would you use SQL to detect anomalous transaction patterns across a dataset of millions of bank connections on Plaid?
