nerdexam
Google

PROFESSIONAL-MACHINE-LEARNING-ENGINEER · Question #113

You have successfully deployed to production a large and complex TensorFlow model trained on tabular data. You want to predict the lifetime value (LTV) field for each subscription stored in the…

The correct answer is B. Add a model monitoring job where 10% of incoming predictions are sampled 24 hours. Vertex AI Model Monitoring detects feature and prediction drift by sampling incoming prediction requests and comparing their distributions to a training baseline. Sampling 10% of predictions every 24 hours is the balanced approach: 10% provides a statistically meaningful sample…

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

Question

You have successfully deployed to production a large and complex TensorFlow model trained on tabular data. You want to predict the lifetime value (LTV) field for each subscription stored in the BigQuery table named subscription. subscriptionPurchase in the project named my-fortune500- company-project. You have organized all your training code, from preprocessing data from the BigQuery table up to deploying the validated model to the Vertex AI endpoint, into a TensorFlow Extended (TFX) pipeline. You want to prevent prediction drift, i.e., a situation when a feature data distribution in production changes significantly over time. What should you do?

Options

  • AImplement continuous retraining of the model daily using Vertex AI Pipelines.
  • BAdd a model monitoring job where 10% of incoming predictions are sampled 24 hours.
  • CAdd a model monitoring job where 90% of incoming predictions are sampled 24 hours.
  • DAdd a model monitoring job where 10% of incoming predictions are sampled every hour.

How the community answered

(68 responses)
  • A
    13% (9)
  • B
    76% (52)
  • C
    3% (2)
  • D
    7% (5)

Explanation

Vertex AI Model Monitoring detects feature and prediction drift by sampling incoming prediction requests and comparing their distributions to a training baseline. Sampling 10% of predictions every 24 hours is the balanced approach: 10% provides a statistically meaningful sample without excessive cost, and a 24-hour window aggregates enough data to detect meaningful distributional shifts without generating alert noise. Option C (90% sampling) is wasteful and unnecessary for drift detection. Option D (every hour) is too frequent for a subscription dataset that changes slowly and would incur unnecessary cost. Option A (daily retraining) responds to drift but does not detect or alert on it-retraining without monitoring blindly refreshes the model without diagnosing whether drift is actually occurring.

Topics

#Model Monitoring#Prediction Drift#Data Drift#MLOps

Community Discussion

No community discussion yet for this question.

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