Great question. Yes, that's a big topic for us. We closely monitor production metrics to detect things like input changes and output drift. This allows us to automatically detect if something changes, so that we can trigger retraining and redeployment.