Machine Learning Engineer Associate · 28% of the exam

Data Preparation for Machine Learning: free practice questions

5 sample questions from our 15-question bank for this domain — answers and explanations included. These are the same scenario-based style as the real AWS exam.

1. A team wants to store, share, and reuse curated features across multiple models and teams, with consistent definitions for training and real-time inference. Which AWS service provides this?

  • A. Amazon SageMaker Feature Store✓ Correct
  • B. Amazon S3 Glacier
  • C. AWS Lambda
  • D. Amazon SNS
Explanation

SageMaker Feature Store is a repository to store, share, and serve curated features consistently for training (offline) and inference (online). S3 Glacier (B) is cold archival storage, Lambda (C) runs code, and SNS (D) is messaging — none is a feature repository.

2. A pipeline ingests high-velocity streaming events (clickstream) that must be captured continuously for near-real-time feature computation. Which AWS service is designed for streaming ingestion?

  • A. Amazon Kinesis Data Streams✓ Correct
  • B. Amazon S3 Glacier Deep Archive
  • C. AWS Artifact
  • D. Amazon Comprehend
Explanation

Amazon Kinesis Data Streams ingests high-throughput streaming data in real time for downstream processing. Glacier Deep Archive (B) is cold storage, Artifact (C) is compliance docs, and Comprehend (D) is an NLP service — none handles streaming ingestion.

3. A team must label thousands of images for a supervised computer vision model, using human labelers with a managed workflow. Which AWS service supports this?

  • A. Amazon SageMaker Ground Truth✓ Correct
  • B. Amazon SageMaker Model Monitor
  • C. AWS Glue DataBrew
  • D. Amazon Athena
Explanation

SageMaker Ground Truth provides managed data-labeling workflows (with human labelers and optional automated labeling). Model Monitor (B) watches production models, Glue DataBrew (C) is visual data prep, and Athena (D) queries S3 with SQL — none is a labeling service.

4. Why is it important to split data into separate TRAINING, VALIDATION, and TEST sets?

  • A. Train the model, tune/select using validation, and get an unbiased final performance estimate on the untouched test set✓ Correct
  • B. To triple the dataset size
  • C. Because AWS bills per split
  • D. To encrypt one third of the data
Explanation

The training set fits the model, the validation set guides tuning/model selection, and the held-out test set gives an unbiased estimate of real-world performance. Option B is false (splitting doesn't add data), and options C and D invent nonexistent reasons.

5. For an ML training job reading millions of records from S3, which combination improves read efficiency and cost?

  • A. Store data as compressed, columnar Parquet and partition it in S3 by a relevant key✓ Correct
  • B. Store one massive uncompressed CSV with no partitioning
  • C. Store each record as a separate tiny JSON object with no compression
  • D. Store the data only in Amazon SNS
Explanation

Compressed columnar Parquet plus sensible S3 partitioning reduces bytes scanned and speeds reads, lowering cost and training time. A single giant uncompressed CSV (B) and millions of tiny objects (C) are both inefficient, and SNS (D) is a messaging service, not storage.

10 more questions in this domain

Practice the full bank with instant grading, flashcards, and a timed mock exam.

Start practicing free
Data Preparation for Machine Learning — Free Machine Learning Engineer Associate Practice Questions | DataCertPrep — Certification Prep