nerdexam
Google

PROFESSIONAL-MACHINE-LEARNING-ENGINEER · Question #154

You have developed an ML model to detect the sentiment of users' posts on your company's social media page to identify outages or bugs. You are using Dataflow to provide real-time predictions on…

The correct answer is A. Deploy the models to a Vertex AI endpoint using the traffic-split=0=80. Vertex AI endpoints natively support deploying multiple model versions simultaneously with a configurable traffic-split parameter, routing a specified percentage of requests to each deployed model. This is a fully managed, first-class feature requiring minimal operational…

Submitted by noor.lb· Apr 18, 2026ML pipeline operationalization

Question

You have developed an ML model to detect the sentiment of users' posts on your company's social media page to identify outages or bugs. You are using Dataflow to provide real-time predictions on data ingested from Pub/Sub. You plan to have multiple training iterations for your model and keep the latest two versions live after every run. You want to split the traffic between the versions in an 80:20 ratio, with the newest model getting the majority of the traffic. You want to keep the pipeline as simple as possible, with minimal management required. What should you do?

Options

  • ADeploy the models to a Vertex AI endpoint using the traffic-split=0=80,
  • BWrap the models inside an App Engine application using the --splits PREVIOUS_VERSION=0.2,
  • CWrap the models inside a Cloud Run container using the REVISION1=20, REVISION2=80
  • DImplement random splitting in Dataflow using beam.Partition() with a partition function calling a

How the community answered

(35 responses)
  • A
    77% (27)
  • B
    3% (1)
  • C
    11% (4)
  • D
    9% (3)

Explanation

Vertex AI endpoints natively support deploying multiple model versions simultaneously with a configurable traffic-split parameter, routing a specified percentage of requests to each deployed model. This is a fully managed, first-class feature requiring minimal operational overhead. Option B (App Engine) and C (Cloud Run) require wrapping models in custom applications and managing traffic splitting manually, adding complexity. Option D (Dataflow with beam.Partition()) adds splitting logic to the streaming pipeline itself, making it more complex and harder to maintain, and it conflates inference infrastructure concerns with data pipeline concerns.

Topics

#ML Model Deployment#Traffic Management#Vertex AI#Real-time Prediction

Community Discussion

No community discussion yet for this question.

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