AIF-C01 Question 46
Single answerA machine learning team has trained a deep learning model to predict customer churn. They now need to deploy the model to production to make predictions accessible to other applications. The team wants to minimize infrastructure management and use a fully managed service. Which deployment method should they use?
- A
Deploy the model using Amazon SageMaker Endpoint.
- B
Host the model on an EC2 instance and expose it via a custom API.
- C
Use AWS Lambda to host the model and expose it as a serverless function.
- D
Deploy the model using Amazon S3 as a static file for inference.
Show answer and explanation
Correct answer: A
Explanation
Amazon SageMaker Endpoint is the most appropriate choice for deploying the model in production with minimal infrastructure management. It is a fully managed service that supports real-time inference, automatic scaling, and integration with other AWS services, making it ideal for this scenario. Other options either require significant infrastructure management or do not support the requirements for dynamic inference.
- A. Correct.
This is the correct option. Amazon SageMaker Endpoints provide a fully managed service for deploying machine learning models, allowing the team to minimize infrastructure management and focus on model inference.
- B. Incorrect.
This is incorrect because hosting the model on an EC2 instance requires significant infrastructure management, including maintaining the instance, scaling, and updating the API.
- C. Incorrect.
This is incorrect because AWS Lambda is not optimized for hosting large machine learning models, particularly deep learning models that require significant computational resources for inference.
- D. Incorrect.
This is incorrect because Amazon S3 is designed for object storage and does not support dynamic inference. It cannot serve as an endpoint for real-time predictions.