AWS Certified Machine Learning - Specialty: Deploying Fraud Prediction Model with SageMaker Random Cut Forest | WebsiteName

Fraud Prediction Model Deployment with SageMaker Random Cut Forest

Question

You work as a machine learning specialist for a banking firm where you are part of the machine learning team in the fraud department.

Your team's latest assignment is to build and deploy a fraud prediction model based on the SageMaker Random Cut Forest built-in algorithm.

You are working through your deployment steps manually using the SageMaker Studio before you automate the pipeline.

You have created an MLOps project in SageMaker Studio and chosen the MLOps template for model building, training, and a deployment project template.

You have cloned the model repo to your local SageMaker Studio environment, made your necessary pipeline changes, committed your code and MLOps has triggered a run of your pipeline.

What are the steps that follow the MLOps triggering of your pipeline?

Answers

Explanations

Click on the arrows to vote for the correct answer

A. B. C. D.

Correct Answer: B.

Option A is incorrect.

You cannot deploy directly to production using the MLOps project flow.

Deploying your SageMaker Studio project through the console requires you to approve the new version, deploy to a staging environment, and use CodePipeline to approve your DeployStaging stage.

Option B is correct.

Deploying your SageMaker Studio project through the console requires you to approve the new version, deploy to a staging environment, and use CodePipeline to approve your DeployStaging stage.

Option C is incorrect.

MLOps first deploys your new DeployStaging stage to your staging environment.

Option D is incorrect.

You are required to approve your new version of your project before it deploys it to staging.

References:

Please see the Amazon SageMaker developer guide titled What is a SageMaker Project? (https://docs.aws.amazon.com/sagemaker/latest/dg/sagemaker-projects-whatis.html),

The Amazon SageMaker developer guide titled SageMaker MLOps Project Walkthrough (https://docs.aws.amazon.com/sagemaker/latest/dg/sagemaker-projects-walkthrough.html)

The correct answer is D.

After MLOps triggers the pipeline, the following steps take place in sequence:

  1. MLOps creates a new model version.

  2. MLOps deploys the new model version to the staging environment. The staging environment is a non-production environment used for testing and validation purposes before the model is deployed to the production endpoint.

  3. Once the model has been deployed to the staging environment, the model's performance is evaluated and tested to ensure that it meets the required performance metrics and that it is stable.

  4. From the CodePipeline, the DeployStaging stage is selected, and the pipeline is approved.

  5. Once the pipeline is approved, the MLOps system deploys the model to the production endpoint. The production endpoint is the environment where the model is used for real-time prediction and fraud detection.

Option A is incorrect because MLOps cannot deploy the model directly to the production endpoint without first deploying it to the staging environment for testing and validation.

Option B is incorrect because after MLOps creates a new model version, the user must approve the new version before it can be deployed to the staging environment. Once it has been approved, it can be deployed to the staging environment, and then the pipeline can be approved to deploy the model to the production endpoint.

Option C is incorrect because MLOps cannot deploy the model directly to the production endpoint without first deploying it to the staging environment for testing and validation.

Therefore, the correct answer is D.