Amazon AWS Certified Machine Learning Engineer - Associate MLA-C01 Practice Test 1

Question 1

A single ML platform team is handling two distinct changes for the same system. Workstream A: An insurance company trains on years of wide structured claims data in Amazon S3, but analysts repeatedly scan only six columns and spend too much time reading CSV files. Workstream B: Several SageMaker training workers require concurrent access to a POSIX directory tree containing many small training files instead of manually syncing copies into every container. Select exactly TWO measures, one that directly resolves each workstream.

Attach a supported EFS filesystem and mount the authorized shared directory in the training environment.
Configure an Amazon Data Firehose stream with an S3 destination, suitable buffering, compression, and failure handling.
Configure an AWS Glue ETL job with an approved JDBC connection and S3 output dataset.
Enable S3 Transfer Acceleration for the eligible bucket and use its accelerated upload endpoint.
Partition the records by date and store compressed Parquet objects in the authorized S3 location.

Explanation

Correct Answers

A:

This action fits the situation: Several SageMaker training workers require concurrent access to a POSIX directory tree containing many small training files instead of manually syncing copies into every container. It implements the needed technical control: Attach a supported EFS filesystem and mount the authorized shared directory in the training environment. The intended outcome is that workers share a consistent POSIX file namespace throughout processing. The important limitation is that eFS throughput and mount access configuration must be sized and restricted for the actual workload.

E:

This action fits the situation: An insurance company trains on years of wide structured claims data in Amazon S3, but analysts repeatedly scan only six columns and spend too much time reading CSV files. It implements the needed technical control: Partition the records by date and store compressed Parquet objects in the authorized S3 location. The intended outcome is that column pruning and compression reduce input read and query cost for the selected fields. The important limitation is that converting to Parquet does not eliminate the need to choose sensible partitions or prevent small-file overhead.

Incorrect Answers

B:

That alternative addresses a different task: A telemetry pipeline receives continuously arriving application events and must land compressed records in S3 without a custom long-running consumer service. The selected workloads instead require Columnar object storage, Shared POSIX training files.

C:

That alternative addresses a different task: A team needs a scheduled extraction of selected records from Amazon RDS into S3 for modeling while applying typed transformations and recording job history. The selected workloads instead require Columnar object storage, Shared POSIX training files.

D:

That alternative addresses a different task: A research partner uploads multi-gigabyte labeled video datasets to a distant AWS Region and the normal S3 transfer route underperforms. The selected workloads instead require Columnar object storage, Shared POSIX training files.

Question 2

A single ML platform team is handling two distinct changes for the same system. Workstream A: A fraud team needs to replay an ordered stream of card authorization events for a new consumer without asking the card processor to send the original events again. Workstream B: A vision team repeatedly reads millions of images from S3 during high-throughput distributed training and requires filesystem semantics with parallel read performance. Select exactly TWO measures, one that directly resolves each workstream.

Choose and provision an EBS volume type and IOPS budget suited to the observed random I/O requirement.
Use Amazon FSx for Lustre with an appropriate S3 data repository association and training access configuration.
Ingest events into Kinesis Data Streams with sufficient shards and retention for consumer replay.
Enable S3 Transfer Acceleration for the eligible bucket and use its accelerated upload endpoint.
Route selected S3 object events through EventBridge to an idempotent validation workflow.

Explanation

Correct Answers

B:

This action fits the situation: A vision team repeatedly reads millions of images from S3 during high-throughput distributed training and requires filesystem semantics with parallel read performance. It implements the needed technical control: Use Amazon FSx for Lustre with an appropriate S3 data repository association and training access configuration. The intended outcome is that distributed workers can read cached or staged training files through a high-performance shared filesystem. The important limitation is that fSx for Lustre incurs filesystem capacity costs and must be managed after the training window.

C:

This action fits the situation: A fraud team needs to replay an ordered stream of card authorization events for a new consumer without asking the card processor to send the original events again. It implements the needed technical control: Ingest events into Kinesis Data Streams with sufficient shards and retention for consumer replay. The intended outcome is that independent consumers can read retained stream records and reprocess events during the retention period. The important limitation is that retention and shard capacity are bounded; a delivery-only destination is not a substitute for replayable streams.

Incorrect Answers

A:

That alternative addresses a different task: A GPU training instance repeatedly reads a random-access local index and standard EBS storage saturates on small I/O requests during preprocessing. The selected workloads instead require Replayable streaming ingestion, High throughput training filesystem.

D:

That alternative addresses a different task: A research partner uploads multi-gigabyte labeled video datasets to a distant AWS Region and the normal S3 transfer route underperforms. The selected workloads instead require Replayable streaming ingestion, High throughput training filesystem.

E:

That alternative addresses a different task: A finance lake receives one S3 object per nightly batch and wants ingestion validation to start when the specific input prefix receives a complete new object. The selected workloads instead require Replayable streaming ingestion, High throughput training filesystem.

Question 3

A single ML platform team is handling two distinct changes for the same system. Workstream A: A telemetry pipeline receives continuously arriving application events and must land compressed records in S3 without a custom long-running consumer service. Workstream B: A data scientist discovers that yesterday's feature extraction changed the raw CSV underlying a model run and must recover the exact original S3 object revision. Select exactly TWO measures, one that directly resolves each workstream.

Enable S3 Versioning and record the exact source object version IDs in the experiment manifest.
Use S3 multipart upload with checkpointed part identifiers and abort rules for abandoned attempts.
Route selected S3 object events through EventBridge to an idempotent validation workflow.
Configure an Amazon Data Firehose stream with an S3 destination, suitable buffering, compression, and failure handling.
Deploy appropriately configured Amazon MSK with secure broker access and existing Kafka-topic conventions.

Explanation

Correct Answers

A:

This action fits the situation: A data scientist discovers that yesterday's feature extraction changed the raw CSV underlying a model run and must recover the exact original S3 object revision. It implements the needed technical control: Enable S3 Versioning and record the exact source object version IDs in the experiment manifest. The intended outcome is that a later experiment can retrieve the same immutable historical object revisions where retained. The important limitation is that versioning does not create a complete data catalog and lifecycle or delete-marker handling must be managed.

D:

This action fits the situation: A telemetry pipeline receives continuously arriving application events and must land compressed records in S3 without a custom long-running consumer service. It implements the needed technical control: Configure an Amazon Data Firehose stream with an S3 destination, suitable buffering, compression, and failure handling. The intended outcome is that buffered streaming records reach the target data lake as durable objects. The important limitation is that firehose buffering adds latency and does not provide arbitrary replay of records already delivered.

Incorrect Answers

B:

That alternative addresses a different task: A remote sensor network periodically uploads 30 GB compressed datasets, and a failed connection currently forces a full restart of each upload. The selected workloads instead require Managed streaming delivery, Versioned raw dataset.

C:

That alternative addresses a different task: A finance lake receives one S3 object per nightly batch and wants ingestion validation to start when the specific input prefix receives a complete new object. The selected workloads instead require Managed streaming delivery, Versioned raw dataset.

E:

That alternative addresses a different task: An enterprise already uses Apache Kafka producers and requires managed Kafka brokers for an AWS-based ML event pipeline without rewriting those clients for Kinesis. The selected workloads instead require Managed streaming delivery, Versioned raw dataset.

Question 4

A single ML platform team is handling two distinct changes for the same system. Workstream A: A research partner uploads multi-gigabyte labeled video datasets to a distant AWS Region and the normal S3 transfer route underperforms. Workstream B: A team needs a scheduled extraction of selected records from Amazon RDS into S3 for modeling while applying typed transformations and recording job history. Select exactly TWO measures, one that directly resolves each workstream.

Partition the records by date and store compressed Parquet objects in the authorized S3 location.
Enable S3 Transfer Acceleration for the eligible bucket and use its accelerated upload endpoint.
Ingest events into Kinesis Data Streams with sufficient shards and retention for consumer replay.
Route selected S3 object events through EventBridge to an idempotent validation workflow.
Configure an AWS Glue ETL job with an approved JDBC connection and S3 output dataset.

Explanation

Correct Answers

B:

This action fits the situation: A research partner uploads multi-gigabyte labeled video datasets to a distant AWS Region and the normal S3 transfer route underperforms. It implements the needed technical control: Enable S3 Transfer Acceleration for the eligible bucket and use its accelerated upload endpoint. The intended outcome is that eligible remote uploads can use the optimized AWS edge network path to reach the bucket. The important limitation is that transfer Acceleration has additional charges and is not a universal speedup for clients near the destination Region.

E:

This action fits the situation: A team needs a scheduled extraction of selected records from Amazon RDS into S3 for modeling while applying typed transformations and recording job history. It implements the needed technical control: Configure an AWS Glue ETL job with an approved JDBC connection and S3 output dataset. The intended outcome is that relational records enter the governed data lake in a repeatable transformation pipeline. The important limitation is that database credentials, VPC connectivity, and read load require explicit protections.

Incorrect Answers

A:

That alternative addresses a different task: An insurance company trains on years of wide structured claims data in Amazon S3, but analysts repeatedly scan only six columns and spend too much time reading CSV files. The selected workloads instead require Long distance object transfer, Relational source extraction.

C:

That alternative addresses a different task: A fraud team needs to replay an ordered stream of card authorization events for a new consumer without asking the card processor to send the original events again. The selected workloads instead require Long distance object transfer, Relational source extraction.

D:

That alternative addresses a different task: A finance lake receives one S3 object per nightly batch and wants ingestion validation to start when the specific input prefix receives a complete new object. The selected workloads instead require Long distance object transfer, Relational source extraction.

Question 5

A single ML platform team is handling two distinct changes for the same system. Workstream A: Several SageMaker training workers require concurrent access to a POSIX directory tree containing many small training files instead of manually syncing copies into every container. Workstream B: A merchant needs downstream features updated when items in a DynamoDB inventory table change, without polling and scanning the entire table each minute. Select exactly TWO measures, one that directly resolves each workstream.

Ingest events into Kinesis Data Streams with sufficient shards and retention for consumer replay.
Use S3 multipart upload with checkpointed part identifiers and abort rules for abandoned attempts.
Enable DynamoDB Streams and consume the relevant change records with an authorized processing integration.
Configure an AWS Glue ETL job with an approved JDBC connection and S3 output dataset.
Attach a supported EFS filesystem and mount the authorized shared directory in the training environment.

Explanation

Correct Answers

C:

This action fits the situation: A merchant needs downstream features updated when items in a DynamoDB inventory table change, without polling and scanning the entire table each minute. It implements the needed technical control: Enable DynamoDB Streams and consume the relevant change records with an authorized processing integration. The intended outcome is that updated item attributes can flow into downstream feature processing with lower polling overhead. The important limitation is that dynamoDB Streams has finite retention and consumers must handle retries and duplicate delivery.

E:

This action fits the situation: Several SageMaker training workers require concurrent access to a POSIX directory tree containing many small training files instead of manually syncing copies into every container. It implements the needed technical control: Attach a supported EFS filesystem and mount the authorized shared directory in the training environment. The intended outcome is that workers share a consistent POSIX file namespace throughout processing. The important limitation is that eFS throughput and mount access configuration must be sized and restricted for the actual workload.

Incorrect Answers

A:

That alternative addresses a different task: A fraud team needs to replay an ordered stream of card authorization events for a new consumer without asking the card processor to send the original events again. The selected workloads instead require Shared POSIX training files, Change data capture events.

B:

That alternative addresses a different task: A remote sensor network periodically uploads 30 GB compressed datasets, and a failed connection currently forces a full restart of each upload. The selected workloads instead require Shared POSIX training files, Change data capture events.

D:

That alternative addresses a different task: A team needs a scheduled extraction of selected records from Amazon RDS into S3 for modeling while applying typed transformations and recording job history. The selected workloads instead require Shared POSIX training files, Change data capture events.

Question 6

An ML engineering team faces this production constraint: A vision team repeatedly reads millions of images from S3 during high-throughput distributed training and requires filesystem semantics with parallel read performance. Which action best addresses the actual requirement?

Use Amazon FSx for Lustre with an appropriate S3 data repository association and training access configuration.
Route selected S3 object events through EventBridge to an idempotent validation workflow.
Enable S3 Transfer Acceleration for the eligible bucket and use its accelerated upload endpoint.
Use S3 multipart upload with checkpointed part identifiers and abort rules for abandoned attempts.

Explanation

Correct Answer

A:

This action fits the situation: A vision team repeatedly reads millions of images from S3 during high-throughput distributed training and requires filesystem semantics with parallel read performance. It implements the needed technical control: Use Amazon FSx for Lustre with an appropriate S3 data repository association and training access configuration. The intended outcome is that distributed workers can read cached or staged training files through a high-performance shared filesystem. The important limitation is that fSx for Lustre incurs filesystem capacity costs and must be managed after the training window.

Incorrect Answers

B:

This action is appropriate for another requirement: A finance lake receives one S3 object per nightly batch and wants ingestion validation to start when the specific input prefix receives a complete new object. It does not solve the presented constraint: A vision team repeatedly reads millions of images from S3 during high-throughput distributed training and requires filesystem semantics with parallel read performance. The matching action for this case is: Use Amazon FSx for Lustre with an appropriate S3 data repository association and training access configuration.

C:

This action is appropriate for another requirement: A research partner uploads multi-gigabyte labeled video datasets to a distant AWS Region and the normal S3 transfer route underperforms. It does not solve the presented constraint: A vision team repeatedly reads millions of images from S3 during high-throughput distributed training and requires filesystem semantics with parallel read performance. The matching action for this case is: Use Amazon FSx for Lustre with an appropriate S3 data repository association and training access configuration.

D:

This action is appropriate for another requirement: A remote sensor network periodically uploads 30 GB compressed datasets, and a failed connection currently forces a full restart of each upload. It does not solve the presented constraint: A vision team repeatedly reads millions of images from S3 during high-throughput distributed training and requires filesystem semantics with parallel read performance. The matching action for this case is: Use Amazon FSx for Lustre with an appropriate S3 data repository association and training access configuration.

Question 7

A cloud ML change request describes the following issue: A data scientist discovers that yesterday's feature extraction changed the raw CSV underlying a model run and must recover the exact original S3 object revision. Which action best addresses the actual requirement?

Use S3 multipart upload with checkpointed part identifiers and abort rules for abandoned attempts.
Enable S3 Versioning and record the exact source object version IDs in the experiment manifest.
Configure an Amazon Data Firehose stream with an S3 destination, suitable buffering, compression, and failure handling.
Configure SageMaker Feature Store with appropriate online and offline storage for shared feature groups.

Explanation

Correct Answer

B:

This action fits the situation: A data scientist discovers that yesterday's feature extraction changed the raw CSV underlying a model run and must recover the exact original S3 object revision. It implements the needed technical control: Enable S3 Versioning and record the exact source object version IDs in the experiment manifest. The intended outcome is that a later experiment can retrieve the same immutable historical object revisions where retained. The important limitation is that versioning does not create a complete data catalog and lifecycle or delete-marker handling must be managed.

Incorrect Answers

A:

This action is appropriate for another requirement: A remote sensor network periodically uploads 30 GB compressed datasets, and a failed connection currently forces a full restart of each upload. It does not solve the presented constraint: A data scientist discovers that yesterday's feature extraction changed the raw CSV underlying a model run and must recover the exact original S3 object revision. The matching action for this case is: Enable S3 Versioning and record the exact source object version IDs in the experiment manifest.

C:

This action is appropriate for another requirement: A telemetry pipeline receives continuously arriving application events and must land compressed records in S3 without a custom long-running consumer service. It does not solve the presented constraint: A data scientist discovers that yesterday's feature extraction changed the raw CSV underlying a model run and must recover the exact original S3 object revision. The matching action for this case is: Enable S3 Versioning and record the exact source object version IDs in the experiment manifest.

D:

This action is appropriate for another requirement: An online fraud API needs current user feature values in milliseconds, while model training needs historical feature records from the same definitions. It does not solve the presented constraint: A data scientist discovers that yesterday's feature extraction changed the raw CSV underlying a model run and must recover the exact original S3 object revision. The matching action for this case is: Enable S3 Versioning and record the exact source object version IDs in the experiment manifest.

Question 8

A team is choosing a specific control for this workload: A team needs a scheduled extraction of selected records from Amazon RDS into S3 for modeling while applying typed transformations and recording job history. Which action best addresses the actual requirement?

Route selected S3 object events through EventBridge to an idempotent validation workflow.
Choose and provision an EBS volume type and IOPS budget suited to the observed random I/O requirement.
Attach a supported EFS filesystem and mount the authorized shared directory in the training environment.
Configure an AWS Glue ETL job with an approved JDBC connection and S3 output dataset.

Explanation

Correct Answer

D:

This action fits the situation: A team needs a scheduled extraction of selected records from Amazon RDS into S3 for modeling while applying typed transformations and recording job history. It implements the needed technical control: Configure an AWS Glue ETL job with an approved JDBC connection and S3 output dataset. The intended outcome is that relational records enter the governed data lake in a repeatable transformation pipeline. The important limitation is that database credentials, VPC connectivity, and read load require explicit protections.

Incorrect Answers

A:

This action is appropriate for another requirement: A finance lake receives one S3 object per nightly batch and wants ingestion validation to start when the specific input prefix receives a complete new object. It does not solve the presented constraint: A team needs a scheduled extraction of selected records from Amazon RDS into S3 for modeling while applying typed transformations and recording job history. The matching action for this case is: Configure an AWS Glue ETL job with an approved JDBC connection and S3 output dataset.

B:

This action is appropriate for another requirement: A GPU training instance repeatedly reads a random-access local index and standard EBS storage saturates on small I/O requests during preprocessing. It does not solve the presented constraint: A team needs a scheduled extraction of selected records from Amazon RDS into S3 for modeling while applying typed transformations and recording job history. The matching action for this case is: Configure an AWS Glue ETL job with an approved JDBC connection and S3 output dataset.

C:

This action is appropriate for another requirement: Several SageMaker training workers require concurrent access to a POSIX directory tree containing many small training files instead of manually syncing copies into every container. It does not solve the presented constraint: A team needs a scheduled extraction of selected records from Amazon RDS into S3 for modeling while applying typed transformations and recording job history. The matching action for this case is: Configure an AWS Glue ETL job with an approved JDBC connection and S3 output dataset.

Question 9

An ML engineering team faces this production constraint: A merchant needs downstream features updated when items in a DynamoDB inventory table change, without polling and scanning the entire table each minute. Which action best addresses the actual requirement?

Ingest events into Kinesis Data Streams with sufficient shards and retention for consumer replay.
Enable DynamoDB Streams and consume the relevant change records with an authorized processing integration.
Use S3 multipart upload with checkpointed part identifiers and abort rules for abandoned attempts.
Configure an AWS Glue ETL job with an approved JDBC connection and S3 output dataset.

Explanation

Correct Answer

B:

This action fits the situation: A merchant needs downstream features updated when items in a DynamoDB inventory table change, without polling and scanning the entire table each minute. It implements the needed technical control: Enable DynamoDB Streams and consume the relevant change records with an authorized processing integration. The intended outcome is that updated item attributes can flow into downstream feature processing with lower polling overhead. The important limitation is that dynamoDB Streams has finite retention and consumers must handle retries and duplicate delivery.

Incorrect Answers

A:

This action is appropriate for another requirement: A fraud team needs to replay an ordered stream of card authorization events for a new consumer without asking the card processor to send the original events again. It does not solve the presented constraint: A merchant needs downstream features updated when items in a DynamoDB inventory table change, without polling and scanning the entire table each minute. The matching action for this case is: Enable DynamoDB Streams and consume the relevant change records with an authorized processing integration.

C:

This action is appropriate for another requirement: A remote sensor network periodically uploads 30 GB compressed datasets, and a failed connection currently forces a full restart of each upload. It does not solve the presented constraint: A merchant needs downstream features updated when items in a DynamoDB inventory table change, without polling and scanning the entire table each minute. The matching action for this case is: Enable DynamoDB Streams and consume the relevant change records with an authorized processing integration.

D:

This action is appropriate for another requirement: A team needs a scheduled extraction of selected records from Amazon RDS into S3 for modeling while applying typed transformations and recording job history. It does not solve the presented constraint: A merchant needs downstream features updated when items in a DynamoDB inventory table change, without polling and scanning the entire table each minute. The matching action for this case is: Enable DynamoDB Streams and consume the relevant change records with an authorized processing integration.

Question 10

A cloud ML change request describes the following issue: An enterprise already uses Apache Kafka producers and requires managed Kafka brokers for an AWS-based ML event pipeline without rewriting those clients for Kinesis. Which action best addresses the actual requirement?

Deploy appropriately configured Amazon MSK with secure broker access and existing Kafka-topic conventions.
Use S3 multipart upload with checkpointed part identifiers and abort rules for abandoned attempts.
Choose and provision an EBS volume type and IOPS budget suited to the observed random I/O requirement.
Route selected S3 object events through EventBridge to an idempotent validation workflow.

Explanation

Correct Answer

A:

This action fits the situation: An enterprise already uses Apache Kafka producers and requires managed Kafka brokers for an AWS-based ML event pipeline without rewriting those clients for Kinesis. It implements the needed technical control: Deploy appropriately configured Amazon MSK with secure broker access and existing Kafka-topic conventions. The intended outcome is that existing Kafka producers can send events to a managed Kafka infrastructure. The important limitation is that mSK brokers still require capacity, partition, authentication, and consumer lag management.

Incorrect Answers

B:

This action is appropriate for another requirement: A remote sensor network periodically uploads 30 GB compressed datasets, and a failed connection currently forces a full restart of each upload. It does not solve the presented constraint: An enterprise already uses Apache Kafka producers and requires managed Kafka brokers for an AWS-based ML event pipeline without rewriting those clients for Kinesis. The matching action for this case is: Deploy appropriately configured Amazon MSK with secure broker access and existing Kafka-topic conventions.

C:

This action is appropriate for another requirement: A GPU training instance repeatedly reads a random-access local index and standard EBS storage saturates on small I/O requests during preprocessing. It does not solve the presented constraint: An enterprise already uses Apache Kafka producers and requires managed Kafka brokers for an AWS-based ML event pipeline without rewriting those clients for Kinesis. The matching action for this case is: Deploy appropriately configured Amazon MSK with secure broker access and existing Kafka-topic conventions.

D:

This action is appropriate for another requirement: A finance lake receives one S3 object per nightly batch and wants ingestion validation to start when the specific input prefix receives a complete new object. It does not solve the presented constraint: An enterprise already uses Apache Kafka producers and requires managed Kafka brokers for an AWS-based ML event pipeline without rewriting those clients for Kinesis. The matching action for this case is: Deploy appropriately configured Amazon MSK with secure broker access and existing Kafka-topic conventions.

Amazon AWS Certified Machine Learning Engineer - Associate MLA-C01 Exam Dumps, Practice Test Questions

Question 1

Explanation

Correct Answers

Incorrect Answers

Question 2

Explanation

Correct Answers

Incorrect Answers

Question 3

Explanation

Correct Answers

Incorrect Answers

Question 4

Explanation

Correct Answers

Incorrect Answers

Question 5

Explanation

Correct Answers

Incorrect Answers

Question 6

Explanation

Correct Answer

Incorrect Answers

Question 7

Explanation

Correct Answer

Incorrect Answers

Question 8

Explanation

Correct Answer

Incorrect Answers

Question 9

Explanation

Correct Answer

Incorrect Answers

Question 10

Explanation

Correct Answer

Incorrect Answers