Exam Professional Machine Learning Engineer topic 1 question 259 discussion - ExamTopics

Problem Description

The question describes a scenario involving an application using a chain of scikit-learn models to predict product pricing. The goal is to deploy this workflow, manage model versions, ensure scalability (including scaling down to zero), minimize resource usage, and reduce manual effort.

Possible Solutions

Four options are provided:

• A: Expose each model as a Vertex AI endpoint and create a custom container endpoint for workflow orchestration.
• B: Create a custom container endpoint loading individual model files and track versions in BigQuery.
• C: Expose each model as a Vertex AI endpoint and use Cloud Run for workflow orchestration.
• D: Load models into Cloud Run and use it for orchestration, tracking versions in BigQuery.

Correct Answer

The suggested answer is C, indicating that using Vertex AI Endpoints for individual models and Cloud Run for workflow orchestration is the best approach.