ML Engineer MLA-C01 Practice Questions & Answers • ExamPrepare

137 questions in this view

Page 1 of 23

Question 1

A company is preparing a dataset to fine-tune a foundation model (FM) on Amazon Bedrock. The source data consists of customer support case metadata that the company stores in Amazon S3 in CSV format. The metadata includes case type, email, full name, and case ID.

An AI engineer needs to transform the data into a supported format for fine-tuning. The AI engineer must mask all personally identifiable information (PII) in the output dataset in Amazon S3. The masked data must support traceability to cases from the same customer without revealing the original customer data.

Which solution will meet these requirements with the LEAST operational overhead?

Create an AWS Glue extract, transform, and load (ETL) job that reads the CSV files from Amazon S3. Enable AWS Glue sensitive data detection. Apply the partial redact action to mask PII values while keeping the first two characters visible. Configure the job to output the data in JSON format to Amazon S3.
Create an AWS Glue extract, transform, and load (ETL) job that reads the CSV files from Amazon S3. Enable AWS Glue sensitive data detection. Apply the cryptographic hash action for the PII entity types. Configure the job to output the data in JSON format to Amazon S3.
Create an AWS Lambda function that reads each CSV record from Amazon S3 and calls Amazon Comprehend DetectPiiEntities. Apply a SHA-256 hash to each detected PII value. Transform the record into JSON format. Write the output to Amazon S3.
Create an AWS Lambda function that reads each CSV record from Amazon S3 and calls Amazon Comprehend DetectPiiEntities. Replace each detected PII value with a partial redaction while keeping the first two characters visible. Transform the record into JSON format. Write the output to Amazon S3.

Select an answer above, then submit your vote. Reading or practicing does not cast a vote.

Question 2

A company is using Amazon SageMaker AI to train a face detection ML model. The company uses the model to ensure that only authorized employees are present on the company premises.

The facial data used to train the model must be encrypted. The company must manage the encryption control. Only authorized employees must be able to view the data.

Which solution will meet these requirements?

Use SageMaker AI IAM roles and policies to control access. Use AWS KMS for encryption.
Use AWS PrivateLink for SageMaker AI to control access. Use AWS KMS for encryption.
Use AWS PrivateLink for SageMaker AI to control access. Use VPC only mode for encryption.
Use SageMaker AI IAM roles and policies to control access. Use AWS CloudTrail for encryption.

Select an answer above, then submit your vote. Reading or practicing does not cast a vote.

Question 3

An AI engineer is designing a Retrieval Augmented Generation (RAG) system for an enterprise knowledge base that contains 500 million document embeddings.

The system has the following distinct access patterns:

A real-time chat-based application that queries only the most recent 30 days of documents
An archival search feature that analysts can use to query historical documents in near real time

The AI engineer must minimize overall storage costs while meeting the latency needs for each access pattern.

Which solution will meet these requirements?

Store all embeddings in an Amazon OpenSearch Serverless vector search collection. Increase the maximum OpenSearch Compute Units (OCUs) quota. Use time-based metadata filters to separate recent and historical data.
Store all embeddings in Amazon RDS for PostgreSQL with the pgvector extension. Create a Hierarchical Navigable Small World (HNSW) index on the embedding column. Use PostgreSQL table partitioning by date to separate recent and historical data.
Store the most recent 30 days of embeddings in an Amazon OpenSearch Serverless vector search collection for the chat-based application. Store the full historical dataset in Amazon S3 Vectors for the archival search feature.
Store all embeddings in Amazon S3 Vectors. Configure the vector index with higher query frequency settings for the chat-based application.

Select an answer above, then submit your vote. Reading or practicing does not cast a vote.

Question 4

An ML engineer is designing a flow in Amazon Bedrock Flows. The ML engineer wants to integrate input prompts as part of the flow. The ML engineer wants to use a prompt template that gives users the ability to customize specific sections of the prompt text. The flow must track changes in the prompt template.

Which solution will meet these requirements with the LEAST operational overhead?

Create an Amazon S3 bucket that has versioning enabled. Create a prompt template that includes the variations. Upload the template to the S3 bucket. Refer to the S3 object template as a storage node in the flow.
Create a parameter in AWS Systems Manager Parameter Store that has versioning enabled. Upload a template definition that includes variations. Create a node in Flows that points to the Parameter Store parameter. Update the node each time a new version of the parameter is created.
Create a prompt template in Amazon Bedrock Prompt Management. Create a prompt node in Flows that refers to the created prompt template. Create a template version in Prompt Management each time the template is modified. Update the reference to the prompt template in Flows.
Create a prompt template as a secret in AWS Secrets Manager. Create an AWS Lambda function that parses the secret and replaces the prompt variant with the corresponding user input. Create a prompt node in Flows that refers to the created secret by retrieving the current version of the secret.

Select an answer above, then submit your vote. Reading or practicing does not cast a vote.

Question 5

An ML engineer is preparing a dataset that contains medical records. The ML engineer will use the dataset to train an ML model to predict the likelihood of patients developing specific diseases. The dataset contains columns for patient ID, age, medical conditions, test results, and a "Disease" target column.

How should the ML engineer configure the data to train the model?

Remove the patient ID column.
Remove the age column.
Remove the medical conditions and test results columns.
Remove the "Disease" target column.

Select an answer above, then submit your vote. Reading or practicing does not cast a vote.

Question 6

A company is deploying a set of agents for rapid experimentation by using Strands Agents. The company will deploy the agents to Amazon Bedrock AgentCore Runtime. The company wants to integrate the deployment as an action in an AWS CodePipeline pipeline.

Which solution will meet these requirements with the LEAST operational overhead?

Create an AWS CodeBuild project that performs an AgentCore Runtime direct code deployment to create the agents. Add an action in the CodePipeline pipeline that runs the CodeBuild project.
Create an AWS CodeDeploy project that includes the correct configurations to deploy the agents. Specify AgentCore Runtime as the target. Add an action in the CodePipeline pipeline that runs the CodeDeploy project.
Create an Amazon ECR repository. Create an AWS CodeBuild project that generates a container image. Configure the project to upload the container image to Amazon ECR and to run AgentCore Runtime by referencing the uploaded image. Add an action in the CodePipeline pipeline that refers to the CodeBuild project.
Create an Amazon ECR repository. Add an action in the CodePipeline pipeline that directly calls AgentCore Runtime to build and deploy the agents.

Select an answer above, then submit your vote. Reading or practicing does not cast a vote.