nerdexam
Google

PROFESSIONAL-MACHINE-LEARNING-ENGINEER · Question #233

You developed a custom model by using Vertex AI to predict your application's user churn rate. You are using Vertex AI Model Monitoring for skew detection. The training data stored in BigQuery…

The correct answer is D. Separate the training dataset into two tables based on demographic and behavioral features. To monitor two models trained on separate feature sets (demographic and behavioral) by splitting traffic and minimizing management, separate the training datasets, and deploy both models to a single Vertex AI endpoint for unified monitoring.

Submitted by yuriko_h· Apr 18, 2026Monitoring, optimizing, and maintaining ML solutions

Question

You developed a custom model by using Vertex AI to predict your application's user churn rate. You are using Vertex AI Model Monitoring for skew detection. The training data stored in BigQuery contains two sets of features - demographic and behavioral. You later discover that two separate models trained on each set perform better than the original model. You need to configure a new model monitoring pipeline that splits traffic among the two models. You want to use the same prediction-sampling-rate and monitoring-frequency for each model. You also want to minimize management effort. What should you do?

Options

  • AKeep the training dataset as is. Deploy the models to two separate endpoints, and submit two
  • BKeep the training dataset as is. Deploy both models to the same endpoint and submit a Vertex AI
  • CSeparate the training dataset into two tables based on demographic and behavioral features.
  • DSeparate the training dataset into two tables based on demographic and behavioral features.

How the community answered

(54 responses)
  • A
    9% (5)
  • B
    4% (2)
  • C
    20% (11)
  • D
    67% (36)

Why each option

To monitor two models trained on separate feature sets (demographic and behavioral) by splitting traffic and minimizing management, separate the training datasets, and deploy both models to a single Vertex AI endpoint for unified monitoring.

AKeep the training dataset as is. Deploy the models to two separate endpoints, and submit two

Deploying models to two separate endpoints would necessitate managing two distinct monitoring jobs, thereby increasing management effort.

BKeep the training dataset as is. Deploy both models to the same endpoint and submit a Vertex AI

Keeping the training dataset as is contradicts the discovery that separate models trained on distinct demographic and behavioral feature sets perform better, requiring separate training data for the new models.

CSeparate the training dataset into two tables based on demographic and behavioral features.

While separating the training dataset is correct, deploying each model to a separate endpoint would increase management effort for monitoring compared to a single endpoint with traffic splitting.

DSeparate the training dataset into two tables based on demographic and behavioral features.Correct

Deploying both models to the same Vertex AI endpoint enables traffic splitting and allows a single Vertex AI Model Monitoring job to observe both, reducing management overhead while maintaining desired monitoring frequency and prediction sampling rates. Separating the training datasets is necessary because the improved models are based on distinct feature sets.

Concept tested: Vertex AI Model Monitoring with multiple models and traffic splitting

Source: https://cloud.google.com/vertex-ai/docs/model-monitoring/overview

Topics

#Vertex AI Model Monitoring#Model Deployment#Data Skew Detection#MLOps Best Practices

Community Discussion

No community discussion yet for this question.

Sign up to join the discussion
Full PROFESSIONAL-MACHINE-LEARNING-ENGINEER Practice