HikeLetter | PMP Training & Certification Online Course | Project Management Learning

1

A business team needs to build a predictive model from tabular data with minimal custom code. What should it evaluate first?

A. A managed AutoML workflow suitable for the data and objective B. A DNS zone C. Cloud NAT D. A VM startup script

2

A team needs to classify customer support text and wants a managed foundation model capability rather than training from scratch.

A. Evaluate a suitable managed generative AI or language-model service B. Build a custom GPU cluster before defining requirements C. Use Cloud CDN as an NLP engine D. Disable data evaluation

3

A business needs a model but has limited labeled examples. What should the ML engineer consider?

A. Evaluate transfer learning, pretrained models, and whether more labeled data is needed B. Assume supervised training will work perfectly without labels C. Use a billing budget as a model D. Ignore validation

4

A team wants to compare a low-code model against a custom model before committing.

A. Define success metrics and compare both approaches on a representative holdout set B. Choose the first model without testing C. Compare only training accuracy D. Ignore inference cost

5

Data scientists need consistent, governed access to shared datasets and model artifacts.

A. Use controlled IAM access, shared metadata, and versioned artifacts B. Email credentials to the whole team C. Make all buckets public D. Use one shared personal account

6

A model developed by one team must be promoted to production by another team.

A. Document the model, data, evaluation, dependencies, and release process B. Transfer only a notebook screenshot C. Remove training metadata D. Deploy without testing

7

A cross-functional team disagrees on model quality. What should it do?

A. Agree on business-aligned evaluation metrics and a common validation dataset B. Use whichever metric looks best C. Compare results on different datasets D. Skip stakeholder review

8

A team needs traceability for experiments and training runs.

A. Track code, data versions, parameters, metrics, and artifacts B. Keep experiment details only in chat C. Overwrite every model file D. Disable logging

9

A notebook prototype performs well but needs repeatable training and deployment.

A. Turn the workflow into a tested, versioned pipeline with managed artifacts B. Run it manually from a personal laptop forever C. Deploy directly without validation D. Delete the feature definitions

10

A model performs well on training data but poorly on unseen data.

A. Investigate overfitting, leakage, data split quality, and regularization B. Increase training accuracy at any cost C. Use the test set repeatedly to tune parameters D. Remove evaluation

Google Cloud Professional Machine Learning Engineer

Overview

Course Description

Course Questions