A deployed model’s inference quality declined after four months. The team wants to be notified if inference quality drops and to prevent recurrence. Which solution meets both goals?

The correct solution is to retrain the model and monitor model drift using Amazon SageMaker Model Monitor. Model drift, a decline in inference quality over time, often requires retraining the model with fresh data to restore performance. Amazon SageMaker Model Monitor is specifically designed to detect and alert on data and model quality issues, including drift, by continuously analyzing predictions and input data. This directly addresses the need for notification and prevention of recurrence. Retraining the model alone (first incorrect option) doesn't provide continuous monitoring. Building a new model (third and fourth incorrect options) is a more drastic and time-consuming step than retraining, and is not always necessary for drift. Amazon SageMaker Feature Store (third incorrect option) is for storing and managing features, not for monitoring model quality. Amazon SageMaker JumpStart (fourth incorrect option) provides pre-built solutions and models, but not a dedicated drift monitoring service.