Free MLA-C02 Practice Questions

Sample questions from our AWS Certified Machine Learning Engineer Exam question bank, grouped by exam domain. Try each one, then open the answer to see the correct option and why every other option is wrong.

Practice the full 200+ question bank

Data Preparation for ML and AI

Question 1

Your project needs to ingest large volumes of structured data in real time, with low latency. You need to decide which AWS service to use for data ingestion, ensuring high throughput and low latency while considering storage costs.

Which of the following services would best optimize your data ingestion process?

Use Amazon Kinesis to ingest streaming data with low latency
Use Amazon S3 Transfer Acceleration to upload large files to S3
Use Amazon RDS for storing structured data with high availability
Use Amazon EFS for scalable file storage with simple access

Show answer and explanation

Correct answer: Use Amazon Kinesis to ingest streaming data with low latency

Why "Use Amazon Kinesis to ingest streaming data with low latency"?

Kinesis is ideal for real-time streaming data with low latency. It is built for high throughput, which is necessary for ingesting large volumes of data in real time.

Incorrect Options:

Use Amazon S3 Transfer Acceleration to upload large files to S3
S3 Transfer Acceleration is for speeding up uploads to S3 but does not provide real-time ingestion.
Use Amazon RDS for storing structured data with high availability
RDS is used for relational databases, but it’s not designed for high-throughput real-time streaming data ingestion.
Use Amazon EFS for scalable file storage with simple access
EFS is a scalable file system, but it is not optimized for real-time data ingestion or low-latency requirements.

Question 2

You have a large, time-series dataset stored in an S3 bucket that you need to preprocess and ingest into Amazon SageMaker Feature Store. The dataset is updated frequently and needs to be ingested in small, incremental batches.

Which method would be the most efficient way to ingest this data into SageMaker Feature Store?

Use Amazon Kinesis to stream the data to SageMaker Feature Store
Use SageMaker Data Wrangler to directly ingest data from S3
Use AWS Glue to run an ETL job and ingest the data into Feature Store
Use AWS Lambda to trigger an ingestion process from S3

Show answer and explanation

Correct answer: Use SageMaker Data Wrangler to directly ingest data from S3

Why "Use SageMaker Data Wrangler to directly ingest data from S3"?

SageMaker Data Wrangler is designed specifically for preprocessing and ingesting structured data into SageMaker services like Feature Store. It allows you to directly ingest data from S3 in a streamlined and efficient way, without the complexity of additional services.

Incorrect Options:

Use Amazon Kinesis to stream the data to SageMaker Feature Store
Kinesis is ideal for real-time streaming, but this is not the best fit for batch-style data ingestion into Feature Store.
Use AWS Glue to run an ETL job and ingest the data into Feature Store
AWS Glue is a powerful ETL tool but adds unnecessary complexity for the relatively simple task of ingesting structured data into Feature Store.
Use AWS Lambda to trigger an ingestion process from S3
Lambda is suitable for event-driven workloads but is more complex and less integrated for direct data ingestion compared to SageMaker Data Wrangler.

Question 3

You are working with a large dataset that requires multiple data cleaning and transformation steps, including handling missing values, detecting outliers, and normalizing numerical features.

Which AWS tool would be most suitable for efficiently performing these tasks?

Use AWS Glue for batch processing and ETL jobs
Use SageMaker Data Wrangler for a streamlined pipeline
Use AWS Glue DataBrew for an easy, visual transformation tool
Use Spark on Amazon EMR for distributed processing

Show answer and explanation

Correct answer: Use SageMaker Data Wrangler for a streamlined pipeline

Why "Use SageMaker Data Wrangler for a streamlined pipeline"?

SageMaker Data Wrangler is specifically designed for data cleaning, transformation, and feature engineering within a machine learning workflow. It provides an intuitive, code-free environment for handling tasks like missing data imputation, outlier detection, and feature normalization.

Incorrect Options:

Use AWS Glue for batch processing and ETL jobs
AWS Glue is more suited for complex ETL pipelines but requires more setup and isn't as tailored to machine learning-specific data preprocessing.
Use AWS Glue DataBrew for an easy, visual transformation tool
DataBrew is a low-code tool for simple data transformations but lacks the depth and flexibility of Data Wrangler for advanced machine learning tasks like feature engineering.
Use Spark on Amazon EMR for distributed processing
While Spark is powerful for distributed data processing, it requires more setup and is not as specialized for machine learning data cleaning and feature transformation as Data Wrangler.

Question 4

What is a crucial factor to consider when classifying data for a machine learning project?

Classifying data based on its sensitivity
Classifying data after model training
Classifying data based on its source
Classifying all data as sensitive

Show answer and explanation

Correct answer: Classifying data based on its sensitivity

Why "Classifying data based on its sensitivity"?

Data classification is a crucial step in data management and protection. It involves categorizing data based on its sensitivity level. This helps in applying appropriate security measures and in compliance with regulations. It is not advisable to classify all data as sensitive as it may lead to unnecessary security measures and costs.

Incorrect Options:

Classifying all data as sensitive
This approach can lead to unnecessary security measures and costs. Not all data is sensitive and requires the same level of protection.
Classifying data based on its source
The source of data is not always an accurate indicator of its sensitivity. Data should be classified based on its content and the level of sensitivity.
Classifying data after model training
Data should be classified before model training. This helps in ensuring that sensitive data is appropriately protected during the model training process.

ML Model and Foundation Model (FM) Development

Question 5

What is a key factor to consider when choosing an AWS artificial intelligence (AI) service to solve a specific business problem?

The cost of the AI service
The popularity of the AI service
The size of the dataset
The specific problem to be solved

Show answer and explanation

Correct answer: The specific problem to be solved

Why "The specific problem to be solved"?

When choosing an AWS AI service, the specific problem to be solved should be the primary consideration. Different AWS AI services are designed to solve different types of problems. For example, Amazon Translate is used for language translation, Amazon Transcribe for speech-to-text conversion, Amazon Rekognition for image and video analysis, and so on.

Incorrect Options:

The cost of the AI service
While cost is a factor to consider, it should not be the primary consideration. The chosen service should be the one that best solves the specific business problem.
The popularity of the AI service
The popularity of an AI service does not necessarily mean it is the best choice for a specific problem. The choice should be based on the specific problem to be solved.
The size of the dataset
While the size of the dataset can influence the choice of AI service, it should not be the primary consideration. The key factor should be the specific problem that needs to be solved.

Question 6

Which of the following is the most critical aspect when selecting an ML model to solve a specific business problem?

The model's accuracy on training data
The model's popularity in the industry
The model's computational efficiency
The model's ability to generalize on unseen data

Show answer and explanation

Correct answer: The model's ability to generalize on unseen data

Why "The model's ability to generalize on unseen data"?

The most critical aspect when selecting an ML model to solve a specific business problem is the model's ability to generalize on unseen data. This is because the ultimate goal of machine learning is to make accurate predictions on new, unseen data, not on the data used for training.

Incorrect Options:

The model's accuracy on training data
While a model's performance on training data is important, it is not the most critical aspect. A model that performs well on training data but poorly on unseen data (a situation known as overfitting) is not useful.
The model's popularity in the industry
The popularity of a model in the industry does not necessarily mean it is the best choice for a specific problem. The choice should be based on the specific problem to be solved, the available data, and the model's ability to generalize on unseen data.
The model's computational efficiency
While computational efficiency is a factor to consider, it should not be the primary consideration. The chosen model should be the one that best solves the specific problem and can generalize well on unseen data.

Question 7

What is a unique advantage of using L1 regularization in a machine learning model?

It induces feature selection
It increases model complexity
It prevents underfitting
It increases training speed

Show answer and explanation

Correct answer: It induces feature selection

Why "It induces feature selection"?

L1 regularization, also known as Lasso regularization, can lead to zero coefficients, which effectively eliminates the weight of the corresponding feature in the model. This means that some of the features are completely neglected for the evaluation of output. So L1 regularization not only helps in reducing overfitting but can help us in feature selection, making it a unique advantage of this regularization technique.

Incorrect Options:

It increases model complexity
Regularization techniques, including L1, are used to prevent overfitting by reducing model complexity, not increasing it.
It prevents underfitting
Regularization techniques are generally used to prevent overfitting, not underfitting. In fact, if the regularization strength is set too high, it may actually cause the model to underfit the data.
It increases training speed
While L1 regularization can reduce the number of features, which might lead to faster training times in some cases, this is not a guaranteed outcome and is not a primary advantage of L1 regularization.

Question 8

Which metric would be most appropriate to use when evaluating a model for a binary classification problem where both classes are equally important and the data is balanced?

Precision
Recall
F1 score
RMSE

Show answer and explanation

Correct answer: F1 score

Why "F1 score"?

The F1 score is the harmonic mean of precision and recall, and it gives equal weight to both these metrics. It is a good choice when you want a balance between precision (how many selected items are relevant) and recall (how many relevant items are selected) and when both classes are equally important. This is especially true in balanced datasets, where false positives and false negatives have similar impact.

Incorrect Options:

Precision
Precision is the ratio of correctly predicted positive observations to the total predicted positives. It is more useful when the costs of false positives are high.
Recall
Recall (Sensitivity) - the ratio of correctly predicted positive observations to the all observations in actual class. It is more useful when the cost of false negatives is high.
RMSE
Root Mean Square Error (RMSE) is not typically used for classification problems. It is a standard way to measure the error of a model in predicting quantitative data.

Deployment and Orchestration of ML and AI Workflows

Question 9

In AWS SageMaker, when is it appropriate to select a multi-model deployment over a multi-container deployment?

When deploying models with different versions
When deploying models that require different resources
When deploying models with different latency requirements
When deploying models that share the same inference code

Show answer and explanation

Correct answer: When deploying models that share the same inference code

Why "When deploying models that share the same inference code"?

In AWS SageMaker, a multi-model deployment is appropriate when you have multiple models that share the same inference code and dependencies. It allows you to deploy multiple models in a single endpoint, which can be more cost-effective and efficient.

Incorrect Options:

When deploying models with different versions
While it's possible to deploy models with different versions in a multi-model deployment, it's not the primary reason to choose a multi-model deployment over a multi-container deployment.
When deploying models that require different resources
If models require different resources, it might be more appropriate to use a multi-container deployment. In a multi-container deployment, each model can have its own container with its own resource configuration.
When deploying models with different latency requirements
If models have different latency requirements, it might be more appropriate to use a multi-container deployment. In a multi-container deployment, each model can have its own container with its own configuration, which can help to meet different latency requirements.

Question 10

Why might you choose to use the 'Bring Your Own Container' (BYOC) approach with SageMaker in your machine learning solution?

BYOC allows for automatic cost optimization
BYOC allows for custom Docker images
BYOC supports all programming languages
BYOC provides built-in auto-scaling

Show answer and explanation

Correct answer: BYOC allows for custom Docker images

Why "BYOC allows for custom Docker images"?

The 'Bring Your Own Container' (BYOC) approach in SageMaker allows you to use a Docker container with your own custom code. This is useful when you need to use specific libraries or frameworks that are not supported by the built-in SageMaker containers, or when you want to have more control over the environment in which your model is running.

Incorrect Options:

BYOC allows for automatic cost optimization
Cost optimization in AWS is not a feature of BYOC. It depends on how you manage your resources (for example, choosing the right instance types, cleaning up unused resources, etc.).
BYOC supports all programming languages
While Docker containers can encapsulate applications written in any language, the support for different languages is a feature of Docker, not of the BYOC approach specifically.
BYOC provides built-in auto-scaling
Auto-scaling is a feature of the specific AWS service (like EC2, ECS, etc.), not of the BYOC approach. You can define auto-scaling policies for your SageMaker endpoints, but the actual scaling is handled by SageMaker, not by the Docker container.

Question 11

When deploying a SageMaker endpoint within a VPC, what is a crucial step to ensure the endpoint can access S3 data?

Set up a VPC endpoint for S3
Enable VPC peering
Use a NAT gateway or instance
Assign a public IP address

Show answer and explanation

Correct answer: Set up a VPC endpoint for S3

Why "Set up a VPC endpoint for S3"?

When deploying a SageMaker endpoint within a VPC, the endpoint does not have direct access to the internet or other AWS services. To allow the endpoint to access S3 data, you need to set up a VPC endpoint for S3. This provides a private connection between your VPC and S3, without requiring access to the internet.

Incorrect Options:

Enable VPC peering
VPC peering is used to connect two VPCs, not to provide access to S3.
Use a NAT gateway or instance
A NAT gateway or instance is used to allow instances in a private subnet to access the internet, not to provide access to S3.
Assign a public IP address
Assigning a public IP address would give the endpoint direct access to the internet, which is not recommended for security reasons.

Question 12

In a CI/CD pipeline for machine learning workflows, which AWS service would be most effective for automating the orchestration of model building and deployment?

AWS CodeBuild
SageMaker Pipelines
AWS CodePipeline
Amazon EventBridge

Show answer and explanation

Correct answer: SageMaker Pipelines

Why "SageMaker Pipelines"?

SageMaker Pipelines is a purpose-built, easy-to-use Continuous Integration and Continuous Deployment (CI/CD) service for machine learning. It provides the orchestration and automation needed in the steps of a machine learning workflow, including data loading, data transformation, model training, model tuning, and model deployment.

Incorrect Options:

AWS CodeBuild
CodeBuild is a fully managed continuous integration service that compiles source code, runs tests, and produces software packages that are ready to deploy. It is not specifically designed for machine learning workflows.
AWS CodePipeline
CodePipeline is a fully managed continuous delivery service that helps you automate your release pipelines for fast and reliable application and infrastructure updates. It is not specifically designed for machine learning workflows.
Amazon EventBridge
EventBridge is a serverless event bus that makes it easy to connect applications together using data from your own applications, integrated Software-as-a-Service (SaaS) applications, and AWS services. It is not specifically designed for machine learning workflows.

Operating, Monitoring, and Securing ML and AI Solutions

Question 13

What is the primary role of monitoring workflows in a machine learning production environment?

To improve model accuracy
To speed up model training
To enhance model interpretability
To detect anomalies in data processing

Show answer and explanation

Correct answer: To detect anomalies in data processing

Why "To detect anomalies in data processing"?

Monitoring workflows in a machine learning production environment is crucial to detect anomalies or errors in data processing or model inference. This ensures the quality of the data being used and the performance of the model.

Incorrect Options:

To improve model accuracy
While monitoring workflows can indirectly lead to improved model accuracy by ensuring data quality and model performance, its primary role is not to improve model accuracy.
To speed up model training
Monitoring workflows does not directly speed up model training. It is used to ensure the quality of data processing and model inference.
To enhance model interpretability
Monitoring workflows does not directly enhance model interpretability. It is used to ensure the quality of data processing and model inference.

Question 14

What is the primary purpose of using SageMaker Clarify in a machine learning workflow?

To speed up model training
To improve model interpretability
To increase model complexity
To detect changes in data distribution

Show answer and explanation

Correct answer: To detect changes in data distribution

Why "To detect changes in data distribution"?

SageMaker Clarify is primarily used to detect changes in the distribution of data that can affect model performance. This is crucial in identifying and addressing drift in ML models.

Incorrect Options:

To speed up model training
While SageMaker Clarify can indirectly speed up model training by ensuring data quality, its primary purpose is not to speed up model training.
To improve model interpretability
SageMaker Clarify does not directly improve model interpretability. Its main function is to monitor data quality and model performance.
To increase model complexity
SageMaker Clarify does not directly increase model complexity. Its main function is to monitor data quality and model performance.

Question 15

Which AWS service is best suited for setting up dashboards to monitor key performance metrics for ML infrastructure?

Amazon QuickSight
Amazon CloudWatch
AWS X-Ray
AWS CloudTrail

Show answer and explanation

Correct answer: Amazon CloudWatch

Why "Amazon CloudWatch"?

Amazon CloudWatch is a monitoring and observability service built for DevOps engineers, developers, site reliability engineers (SREs), and IT managers. CloudWatch provides you with data and actionable insights to monitor your applications, understand and respond to system-wide performance changes, optimize resource utilization, and get a unified view of operational health.

Incorrect Options:

Amazon QuickSight
Amazon QuickSight is a fast, cloud-powered business intelligence service that makes it easy to deliver insights to everyone in your organization. While it can be used to create dashboards, it is not specifically designed for monitoring key performance metrics for ML infrastructure.
AWS X-Ray
AWS X-Ray helps developers analyze and debug production, distributed applications, such as those built using a microservices architecture. While it provides insights into the behavior of your applications, it is not primarily used for setting up dashboards to monitor performance metrics.
AWS CloudTrail
AWS CloudTrail is a service that enables governance, compliance, operational auditing, and risk auditing of your AWS account. It is primarily used to log, monitor, and invoke re-training activities for machine learning models and not for setting up dashboards to monitor performance metrics.

Question 16

What is the best practice when troubleshooting and debugging security issues in a CI/CD pipeline for a Machine Learning project on AWS?

Ignore minor security alerts
Use AWS X-Ray for deep dive diagnostics
Disable security controls for easier debugging
Avoid using AWS provided tools

Show answer and explanation

Correct answer: Use AWS X-Ray for deep dive diagnostics

Why "Use AWS X-Ray for deep dive diagnostics"?

AWS X-Ray helps developers analyze and debug production, distributed applications, such as those built using a microservices architecture. With X-Ray, you can understand how your application and its underlying services are performing to identify and troubleshoot the root cause of performance issues and errors. It provides an end-to-end view of requests as they travel through your application, and shows a map of your application’s underlying components.

Incorrect Options:

Ignore minor security alerts
Ignoring any security alerts, even minor ones, can lead to larger security issues. It's important to investigate all security alerts.
Disable security controls for easier debugging
Disabling security controls can expose your application to potential threats. It's not a recommended practice.
Avoid using AWS provided tools
AWS provides a number of tools that are specifically designed to help with troubleshooting and debugging. Avoiding these tools can make the process more difficult and less effective.

Ready for the full MLA-C02 practice exam?

Practice the full 200+ question bank