Answer: Amazon SageMaker Model Monitor

## Detailed Explanation This question involves two key requirements for a media company: 1. **Deploying a customized ML model** for analyzing viewer behavior and demographics to provide personalized content recommendations 2. **Monitoring the deployed model for quality drift over time** Let's analyze each option: **A: Amazon Rekognition** - This is a computer vision service for image and video analysis. While it could potentially analyze visual content, it's not designed for deploying custom ML models for behavioral analysis or monitoring model drift. It's primarily a pre-trained service for specific computer vision tasks. **B: Amazon SageMaker Clarify** - This service helps detect bias in ML models and explain predictions. While it's valuable for model interpretability and fairness, it's not primarily designed for continuous monitoring of model quality drift in production environments. **C: Amazon Comprehend** - This is a natural language processing service for text analysis. It could analyze text-based viewer feedback or content, but it's not designed for deploying custom ML models or monitoring model drift over time. **D: Amazon SageMaker Model Monitor** - This is the correct choice because: - **SageMaker** provides the complete platform for building, training, and deploying custom ML models, which addresses the first requirement of deploying a customized model for viewer behavior analysis - **Model Monitor** is specifically designed to continuously monitor ML models in production for quality drift, which directly addresses the second requirement - The service automatically detects concept drift (when relationships between input data and predictions change) and data drift (when input data distribution changes over time) - It provides alerts and visualizations to help identify when model performance degrades, which is crucial for maintaining accurate personalized recommendations as viewer behavior evolves For a media company analyzing viewer behavior and demographics, SageMaker Model Monitor would enable them to: 1. Build and deploy a custom recommendation model using SageMaker 2. Continuously monitor that model's performance in production 3. Detect when the model's predictions become less accurate due to changing viewer preferences or demographics 4. Take corrective actions (like retraining the model) when drift is detected This combination of model deployment and continuous monitoring ensures that personalized content recommendations remain relevant and accurate over time as viewer behavior patterns change.

Author: LeetQuiz Editorial Team