MLA-C01 Question 15
Select 3You are tasked with building a machine learning model for predicting customer churn using Amazon SageMaker. After training multiple models, you decide to deploy the best-performing model to an endpoint for real-time inference. Which of the following steps should you take to ensure the deployed model performs optimally and securely in production?
- A
Enable Amazon CloudWatch monitoring for the endpoint to track inference performance and errors.
- B
Use SageMaker Model Monitor to detect data drift and ensure the input data distribution matches the training data.
- C
Store sensitive customer data directly in the model artifacts to improve model performance.
- D
Configure Amazon SageMaker endpoint auto-scaling to handle variable traffic loads.
- E
Deploy the model using an unencrypted endpoint to improve latency.
Show answer and explanation
Correct answers: A, B, D
Explanation
To ensure optimal and secure deployment of a machine learning model in production, it is important to monitor the endpoint's performance, detect data drift with SageMaker Model Monitor, and configure auto-scaling to handle varying traffic. Security best practices, such as avoiding storing sensitive data in model artifacts and using encrypted endpoints, must also be followed to maintain compliance and protect customer data.
- A. Correct.
Correct. Monitoring inference performance and errors using Amazon CloudWatch will help you identify issues like latency spikes or failures in endpoint invocations.
- B. Correct.
Correct. SageMaker Model Monitor allows you to detect data drift, ensuring that the data the model sees in production is consistent with the data it was trained on.
- C. Incorrect.
Incorrect. Storing sensitive customer data in the model artifacts is a security violation and does not improve performance. Sensitive data should be securely handled according to compliance standards.
- D. Correct.
Correct. Configuring endpoint auto-scaling ensures that your model can handle varying levels of traffic without affecting performance.
- E. Incorrect.
Incorrect. Deploying an unencrypted endpoint compromises security and does not guarantee improved latency. Encryption should always be enabled for production endpoints for data security.