1
AWS Certified Machine
Learning Engineer –
Associate (MLA-C01)
Comprehensive Practice
Exam v2.0 — 150
Multiple-Choice Questions
a well detailed one 2025
/ 2026 written and
graded A+ upgraded
, 2
Exam Overview: The AWS Certified Machine Learning Engineer – Associate (MLA-C01) replaced
the Machine Learning Specialty (MLS-C01) after March 31, 2026. The new exam emphasizes
MLOps, GenAI integration, cost-aware deployment, and production ML engineering over deep
algorithmic theory. The exam consists of 65 questions (50 scored, 15 unscored) over 130
minutes. Candidates should have at least 1 year of experience with SageMaker and ML
engineering.
Exam Domains:
• Domain 1: Data Preparation for Machine Learning — 28%
• Domain 2: ML Model Development — 26%
• Domain 3: Deployment and Orchestration of ML Workflows — 22%
• Domain 4: ML Solution Monitoring, Maintenance, and Security — 24%
DOMAIN 1: DATA PREPARATION FOR MACHINE LEARNING (28%)
Questions 1–42
Question 1
A Machine Learning Engineer is building a data pipeline that ingests 2 TB of daily JSON logs from
multiple sources into Amazon S3. The data must be converted to Apache Iceberg format with
ACID transaction support, time travel capabilities, and schema evolution. The pipeline must run
, 3
daily and complete within 90 minutes. Which combination of AWS services should the Engineer
use?
A. AWS Glue with Spark and S3 Tables (Iceberg)
B. Amazon EMR with Spark and Apache Hive
C. AWS Glue with Python Shell and Parquet
D. Amazon Athena with CTAS statements
- detailed answer 100% correct :- A
Rationale: AWS Glue with Spark supports Apache Iceberg via S3 Tables, providing ACID
transactions, time travel, and schema evolution. Option B requires cluster management. Option
C (Python Shell) lacks the scale for 2 TB. Option D (Athena CTAS) is for query-based table
creation, not ETL pipelines.
Question 2
A financial services company stores sensitive customer data in Amazon S3 for ML training. The
data includes PII such as names, addresses, and Social Security numbers. The company must
automatically detect and redact PII before the data is used for model training. Which AWS
service should the Engineer use?
A. Amazon Macie with automated sensitive data discovery
B. AWS Glue DataBrew with PII detection
C. SageMaker Data Wrangler with built-in transforms
D. Amazon Comprehend with PII detection
- detailed answer 100% correct :- A
Rationale: Amazon Macie is designed for sensitive data discovery in S3 and can automatically
detect PII. Glue DataBrew and SageMaker Data Wrangler are for data transformation, not
security scanning. Comprehend is for text analysis, not S3 data discovery.
Question 3
A Machine Learning Engineer needs to create a feature store that supports both low-latency
online inference (<5 ms) and batch retrieval for offline training. The feature store must serve
features to SageMaker endpoints and training jobs. Which AWS service should the Engineer
use?
, 4
A. SageMaker Feature Store with Online and Offline stores
B. Amazon DynamoDB with S3 backup
C. Amazon ElastiCache with persistence
D. Amazon Redshift with Spectrum
- detailed answer 100% correct :- A
Rationale: SageMaker Feature Store provides both Online Store (low-latency, <5 ms) and Offline
Store (batch retrieval). It integrates natively with SageMaker endpoints and training. DynamoDB
lacks feature versioning and offline store capabilities.
Question 4
A team ingests streaming data from 10,000 IoT devices using Amazon Kinesis Data Streams. The
data arrives as semi-structured JSON and must be validated, transformed, and delivered to S3 in
Parquet format with hourly partitioning. The transformation includes simple filtering and
enrichment. Which service should the team use?
A. Kinesis Data Firehose with built-in Parquet conversion and Lambda transformation
B. Kinesis Data Analytics for Apache Flink
C. AWS Glue streaming ETL
D. Amazon MSK with Kafka Connect
- detailed answer 100% correct :- A
Rationale: Kinesis Data Firehose is the fully managed service for streaming ingestion with built-
in Parquet/ORC conversion and partitioning. Lambda transformations handle simple filtering
and enrichment. Flink is overkill for simple transformations.
Question 5
A Machine Learning Engineer is preparing a dataset with 500 features and 50,000 rows. The
target variable is binary (fraud vs. non-fraud). The Engineer notices that 300 features have
missing values (>30% missing) and 50 features have zero variance. What should the Engineer do
first?
A. Remove features with >30% missing values and zero-variance features
B. Impute missing values with the mean for all features
C. Apply PCA to reduce dimensionality
D. Use SMOTE to handle class imbalance
AWS Certified Machine
Learning Engineer –
Associate (MLA-C01)
Comprehensive Practice
Exam v2.0 — 150
Multiple-Choice Questions
a well detailed one 2025
/ 2026 written and
graded A+ upgraded
, 2
Exam Overview: The AWS Certified Machine Learning Engineer – Associate (MLA-C01) replaced
the Machine Learning Specialty (MLS-C01) after March 31, 2026. The new exam emphasizes
MLOps, GenAI integration, cost-aware deployment, and production ML engineering over deep
algorithmic theory. The exam consists of 65 questions (50 scored, 15 unscored) over 130
minutes. Candidates should have at least 1 year of experience with SageMaker and ML
engineering.
Exam Domains:
• Domain 1: Data Preparation for Machine Learning — 28%
• Domain 2: ML Model Development — 26%
• Domain 3: Deployment and Orchestration of ML Workflows — 22%
• Domain 4: ML Solution Monitoring, Maintenance, and Security — 24%
DOMAIN 1: DATA PREPARATION FOR MACHINE LEARNING (28%)
Questions 1–42
Question 1
A Machine Learning Engineer is building a data pipeline that ingests 2 TB of daily JSON logs from
multiple sources into Amazon S3. The data must be converted to Apache Iceberg format with
ACID transaction support, time travel capabilities, and schema evolution. The pipeline must run
, 3
daily and complete within 90 minutes. Which combination of AWS services should the Engineer
use?
A. AWS Glue with Spark and S3 Tables (Iceberg)
B. Amazon EMR with Spark and Apache Hive
C. AWS Glue with Python Shell and Parquet
D. Amazon Athena with CTAS statements
- detailed answer 100% correct :- A
Rationale: AWS Glue with Spark supports Apache Iceberg via S3 Tables, providing ACID
transactions, time travel, and schema evolution. Option B requires cluster management. Option
C (Python Shell) lacks the scale for 2 TB. Option D (Athena CTAS) is for query-based table
creation, not ETL pipelines.
Question 2
A financial services company stores sensitive customer data in Amazon S3 for ML training. The
data includes PII such as names, addresses, and Social Security numbers. The company must
automatically detect and redact PII before the data is used for model training. Which AWS
service should the Engineer use?
A. Amazon Macie with automated sensitive data discovery
B. AWS Glue DataBrew with PII detection
C. SageMaker Data Wrangler with built-in transforms
D. Amazon Comprehend with PII detection
- detailed answer 100% correct :- A
Rationale: Amazon Macie is designed for sensitive data discovery in S3 and can automatically
detect PII. Glue DataBrew and SageMaker Data Wrangler are for data transformation, not
security scanning. Comprehend is for text analysis, not S3 data discovery.
Question 3
A Machine Learning Engineer needs to create a feature store that supports both low-latency
online inference (<5 ms) and batch retrieval for offline training. The feature store must serve
features to SageMaker endpoints and training jobs. Which AWS service should the Engineer
use?
, 4
A. SageMaker Feature Store with Online and Offline stores
B. Amazon DynamoDB with S3 backup
C. Amazon ElastiCache with persistence
D. Amazon Redshift with Spectrum
- detailed answer 100% correct :- A
Rationale: SageMaker Feature Store provides both Online Store (low-latency, <5 ms) and Offline
Store (batch retrieval). It integrates natively with SageMaker endpoints and training. DynamoDB
lacks feature versioning and offline store capabilities.
Question 4
A team ingests streaming data from 10,000 IoT devices using Amazon Kinesis Data Streams. The
data arrives as semi-structured JSON and must be validated, transformed, and delivered to S3 in
Parquet format with hourly partitioning. The transformation includes simple filtering and
enrichment. Which service should the team use?
A. Kinesis Data Firehose with built-in Parquet conversion and Lambda transformation
B. Kinesis Data Analytics for Apache Flink
C. AWS Glue streaming ETL
D. Amazon MSK with Kafka Connect
- detailed answer 100% correct :- A
Rationale: Kinesis Data Firehose is the fully managed service for streaming ingestion with built-
in Parquet/ORC conversion and partitioning. Lambda transformations handle simple filtering
and enrichment. Flink is overkill for simple transformations.
Question 5
A Machine Learning Engineer is preparing a dataset with 500 features and 50,000 rows. The
target variable is binary (fraud vs. non-fraud). The Engineer notices that 300 features have
missing values (>30% missing) and 50 features have zero variance. What should the Engineer do
first?
A. Remove features with >30% missing values and zero-variance features
B. Impute missing values with the mean for all features
C. Apply PCA to reduce dimensionality
D. Use SMOTE to handle class imbalance
