A startup has a custom PyTorch model with a non-standard preprocessing step that requires a specific library. They want to deploy it to Vertex AI Endpoints with minimal changes to their serving code. What should they do?

Package the model and preprocessing code in a custom container and deploy it to a Vertex AI endpoint

Deploy the model to Cloud Run and call it from Vertex AI Endpoints

Use the Vertex AI prebuilt PyTorch container and add the library via a custom dependency install script

Convert the model to TensorFlow SavedModel and use the prebuilt TensorFlow serving container

When deploying a trained model to an endpoint, what must you specify?

The feature store configuration

The source code of the training pipeline

The model artifact and its serving container or runtime

The training dataset and hyperparameters

A team serves a large image-classification model on a GPU endpoint. The model has 50M parameters and achieves 95% accuracy. The team wants to reduce serving cost by 50% while keeping accuracy above 90%. They have tried quantization, but accuracy dropped to 88%. What should they do next?

Use a more aggressive quantization scheme like int4

Apply knowledge distillation to train a smaller student model that mimics the larger teacher model

Prune the model by removing 50% of the weights

Switch to a smaller GPU machine type without changing the model

A team deploys a model to a Vertex AI endpoint with autoscaling enabled. They notice that during a traffic spike, the endpoint scales up but new requests fail with 429 errors for several seconds. What is the most likely cause and solution?

The model is too large; use a smaller model

The endpoint has too many replicas; reduce the maxReplicaCount

The endpoint is using a private endpoint; switch to a public endpoint

The autoscaling policy is too slow to react; set a lower CPU utilization target to trigger scale-out earlier

Google CloudProfessional Machine Learning Engineer

Domain 4Objective 2

Scaling Online Model Serving PROFESSIONAL-MACHINE-LEARNING-ENGINEER Practice Questions (Page 3)

Part of the Serving and scaling models domain, which accounts for ~20% of the PROFESSIONAL-MACHINE-LEARNING-ENGINEER exam.

37questions here

8free pages

14concepts

~20%of the exam

14 Concepts in This Objective

Feature Store fundamentals
Serving features from Feature Store
Feature Store configuration and monitoring
Public and private endpoints
Deploying models to endpoints
Endpoint security and access control
Hardware selection for serving
Hardware optimization and cost trade-offs
Scaling serving infrastructure
Containerized serving and inference platforms
Throughput and latency management
Model tuning for production
Serving-specific model optimization
Production model lifecycle management

Questions 11–15

11application · medium
A startup has a custom PyTorch model with a non-standard preprocessing step that requires a specific library. They want to deploy it to Vertex AI Endpoints with minimal changes to their serving code. What should they do?
Select an answer first
12foundation · easy
When deploying a trained model to an endpoint, what must you specify?
Select an answer first
13foundation · easy
Which Google Cloud service can be used to restrict access to a model endpoint based on IAM roles?
Select an answer first
14expert · hard
A team serves a large image-classification model on a GPU endpoint. The model has 50M parameters and achieves 95% accuracy. The team wants to reduce serving cost by 50% while keeping accuracy above 90%. They have tried quantization, but accuracy dropped to 88%. What should they do next?
Select an answer first
15application · medium
A team deploys a model to a Vertex AI endpoint with autoscaling enabled. They notice that during a traffic spike, the endpoint scales up but new requests fail with 429 errors for several seconds. What is the most likely cause and solution?
Select an answer first

Finished these 5 questions?

Review the revealed explanations, or continue through the curriculum.

Next 5 questions →Back to PROFESSIONAL-MACHINE-LEARNING-ENGINEER

Free Basic Practice is a study aid with revealable answers — not a scored exam. Examers.io is independent and not affiliated with or endorsed by Google Cloud. “PROFESSIONAL-MACHINE-LEARNING-ENGINEER” is a trademark of its owner, used for identification only.