An ML engineer needs to train a supervised deep learning model. The available dataset is a large number of unlabeled images that only employees should access. The ML engineer needs to implement a solution that labels the dataset with the highest possible accuracy.
Which combination of steps should the ML engineer take to meet these requirements? (Choose two.)
A. Use Amazon Rekognition to automatically label the dataset.
B. Train the deep learning model directly on the raw data. Let the model infer the labels by itself.
C. Use Amazon SageMaker Ground Truth to create an annotation job that specifies the labeling task and requirements.
D. Set up workforce teams to access a private workforce to run and review the annotation job created by Amazon SageMaker Ground Truth.
E. Use Amazon Mechanical Turk to complete the annotation job created by Amazon SageMaker Ground Truth.
Show Answer
Correct Answer: C, D
Explanation: Supervised deep learning requires labeled training data. Amazon SageMaker Ground Truth is the AWS service designed to create and manage data labeling jobs. Because the images should only be accessed by employees, the labeling job should use a private workforce rather than a public workforce such as Mechanical Turk. Amazon Rekognition can assist with auto-labeling in some cases but does not satisfy the requirement for the highest possible accuracy on a sensitive dataset, and training directly on unlabeled data is not supervised learning.
Question 103
An ML engineer needs to deploy a trained model that is based on a genetic algorithm. The algorithm solves a complex problem and can take several minutes to generate predictions.
When the model is deployed, the model needs to access large amounts of data to process requests. The requests can involve as much as 100 MB of data.
Which deployment solution will meet these requirements with the LEAST operational overhead?
A. Deploy the model to Amazon EC2 instances in an Auto Scaling group behind an Application Load Balancer.
B. Deploy the model to an Amazon SageMaker real-time endpoint.
C. Deploy the model to an Amazon SageMaker Asynchronous Inference endpoint.
D. Package the model as a container. Deploy the model to Amazon Elastic Container Service (Amazon ECS) on Amazon EC2 instances.
Show Answer
Correct Answer: C
Explanation: Amazon SageMaker Asynchronous Inference is designed for inference requests that can take minutes to complete and supports large payloads (up to hundreds of MB). It queues requests, processes them asynchronously, and stores responses in Amazon S3, minimizing operational overhead compared with managing EC2 or ECS infrastructure. SageMaker real-time endpoints are intended for low-latency inference, while EC2/ECS deployments require more infrastructure management.
Question 104
A company has an existing Amazon SageMaker model (v1) on a production endpoint. The company develops a new model version (v2) and needs to test v2 in production before substituting v2 for v1.
The company needs to implement a solution to minimize the risk of v2 generating incorrect output in production. The solution must prevent any disruption of production traffic during the change to v2.
Which solution will meet these requirements?
A. Create a second production variant for v2. Assign 1% of the traffic to v2 and 99% of the traffic to v1. Collect all the output of v2 in an Amazon S3 bucket. If v2 performs as expected, switch all the traffic to v2.
B. Create a second production variant for v2. Assign 10% of the traffic to v2 and 90% of the traffic to v1. Collect all the output of v2 in an Amazon S3 bucket. If v2 performs as expected, switch all the traffic to v2.
C. Deploy v2 to a new endpoint. Turn on data capturing for the production endpoint. Write a script to pass 100% of input data to v2. If v2 performs as expected, deactivate the v1 endpoint and direct the traffic to v2.
D. Deploy v2 into a shadow variant that samples 100% of the inference requests. Collect all the output in an Amazon S3 bucket. If v2 performs as expected, promote v2 to production.
Show Answer
Correct Answer: D
Explanation: Shadow variants in Amazon SageMaker allow a new model version to receive a copy of production inference requests without affecting live responses. Production traffic continues to be served by v1, eliminating disruption while collecting v2 outputs for evaluation. After validation, the shadow model can be promoted to production. Canary traffic splitting (1% or 10%) exposes some real users to v2 responses, and deploying to a separate endpoint with custom duplication is more complex and does not match the built-in zero-impact shadow testing capability.
Question 105
A company runs an ML model on Amazon SageMaker. The company uses an automatic process that makes API calls to create training jobs for the model. The company has new compliance rules that prohibit the collection of aggregated metadata from training jobs.
Which solution will prevent SageMaker from collecting metadata from the training jobs?
A. Opt out of metadata tracking for any training job that is submitted.
B. Ensure that training jobs are running in a private subnet in a custom VPC.
C. Encrypt the training data with an AWS Key Management Service (AWS KMS) customer managed key.
D. Reconfigure the training jobs to use only AWS Nitro instances.
Show Answer
Correct Answer: A
Explanation: Amazon SageMaker allows customers to opt out of collecting and storing aggregated training job metadata for submitted training jobs. The other options (private VPC, KMS encryption, or Nitro instances) improve networking, encryption, or infrastructure security but do not disable SageMaker's metadata collection.
Question 106
A company is planning to use an Amazon SageMaker prebuilt algorithm to create a recommendation model. The algorithm must be able to make predictions on high-dimensional sparse data.
Which SageMaker algorithm should the company choose for the recommendation model?
A. K-nearest neighbors (k-NN)
B. Factorization Machines
C. Principal component analysis (PCA)
D. Sequence-to-Sequence (seq2seq)
Show Answer
Correct Answer: B
Explanation: Factorization Machines are specifically designed for recommendation tasks on high-dimensional sparse datasets. They efficiently model pairwise feature interactions using latent factors, making them well suited for user-item recommendation problems. k-NN is a general classification/regression algorithm, PCA is for dimensionality reduction, and seq2seq is for sequence transformation tasks.
Question 107
A company is using an Amazon S3 bucket to collect data that will be used for ML workflows. The company needs to use AWS Glue DataBrew to clean and normalize the data.
Which solution will meet these requirements?
A. Create a DataBrew dataset by using the S3 path. Clean and normalize the data by using a DataBrew profile job.
B. Create a DataBrew dataset by using the S3 path. Clean and normalize the data by using a DataBrew recipe job.
C. Create a DataBrew dataset by using a Java Database Connectivity (JDBC) driver to connect to the S3 bucket. Clean and normalize the data by using a DataBrew profile job.
D. Create a DataBrew dataset by using a Java Database Connectivity (JDBC) driver to connect to the S3 bucket. Clean and normalize the data by using a DataBrew recipe job.
Show Answer
Correct Answer: B
Explanation: AWS Glue DataBrew creates datasets directly from Amazon S3 using the S3 path. To clean and normalize data, you use a DataBrew recipe job, which applies transformation steps and writes the transformed output. A profile job is used for data profiling and analysis rather than transformation. JDBC is not used to connect to Amazon S3.
Question 108
An ML engineer needs to deploy four ML models in an Amazon SageMaker inference pipeline. The models were built with different frameworks. The ML engineer also needs to give clients the ability to use the invoke_endpoint call to perform inference for each model.
Which solution will meet these requirements MOST cost-effectively?
A. Create a SageMaker multi-model endpoint.
B. Create a SageMaker multi-container endpoint.
C. Create multiple SageMaker single-model endpoints.
D. Run a SparkML job to generate multiple endpoints.
Show Answer
Correct Answer: B
Explanation: A SageMaker multi-container endpoint allows multiple models that use different inference frameworks or containers to be deployed behind a single endpoint, making it more cost-effective than separate endpoints. Multi-model endpoints require models to share the same serving container/framework, so they are not suitable when the models are built with different frameworks. Clients can invoke the endpoint and route requests to the appropriate container for inference.
Question 109
An ML engineer is developing a classification model. The ML engineer needs to use custom libraries in processing jobs, training jobs, and pipelines in Amazon SageMaker.
Which solution will provide this functionality with the LEAST implementation effort?
A. Manually install the libraries in the SageMaker containers.
B. Build a custom Docker container that includes the required libraries. Host the container in Amazon Elastic Container Registry (Amazon ECR). Use the ECR image in the SageMaker jobs and pipelines.
C. Create a SageMaker notebook instance to host the jobs. Create an AWS Lambda function to install the libraries on the notebook instance when the notebook instance starts. Configure the SageMaker jobs and pipelines to run on the notebook instance.
D. Run code for the libraries externally on Amazon EC2 instances. Store the results in Amazon S3. Import the results into the SageMaker jobs and pipelines.
Show Answer
Correct Answer: B
Explanation: A custom Docker image stored in Amazon ECR is the standard SageMaker approach for packaging custom dependencies so they are consistently available across processing jobs, training jobs, and pipelines. It is reusable, integrates natively with SageMaker, and avoids repeatedly installing libraries or relying on external infrastructure.
Question 110
A company wants to use Amazon SageMaker to host an ML model that runs on CPU for real-time predictions. The model will have intermittent traffic during business hours and will have periods of no traffic after business hours. The company needs a solution that will serve inference requests in the most cost-effective manner.
Which hosting option will meet these requirements?
A. Deploy the model to a SageMaker real-time endpoint. Add a schedule-based auto scaling policy to handle traffic surges during business hours.
B. Deploy the model to a SageMaker Serverless Inference endpoint. Configure increased provisioned concurrency during business hours.
C. Deploy the model to a SageMaker Asynchronous Inference endpoint. Configure an auto scaling policy that scales in to zero outside business hours.
D. Deploy the model to a SageMaker real-time endpoint. Create a scheduled AWS Lambda function that activates the endpoint during business hours only.
Show Answer
Correct Answer: B
Explanation: SageMaker Serverless Inference is designed for workloads with intermittent or unpredictable traffic. It automatically scales down when idle, minimizing costs, and provisioned concurrency can be scheduled during business hours to reduce cold starts while maintaining responsiveness. A real-time endpoint incurs instance costs even when idle, asynchronous inference is intended for long-running/non-real-time workloads, and stopping/starting real-time endpoints via Lambda is not the intended cost-optimized pattern for real-time intermittent traffic.
Question 111
A company wants to provide services to help other businesses label images. The company wants its labeling specialists to complete human labeling tasks on AWS.
How should the company register the labeling specialists to receive tasks on AWS?
A. Use AWS Data Exchange.
B. Create and use an internal workforce in Amazon SageMaker Ground Truth.
C. Create and use Amazon Mechanical Turk entities in an Amazon SageMaker human loop.
D. Use the Amazon Mechanical Turk website.
Show Answer
Correct Answer: B
Explanation: Amazon SageMaker Ground Truth supports private (internal) workforces for human labeling tasks. A company can register its own labeling specialists as an internal workforce and assign labeling jobs to them through Ground Truth. AWS Data Exchange is unrelated to labeling, Mechanical Turk entities in a SageMaker human loop are for public/crowd workers rather than registering a company's own workforce, and the Mechanical Turk website is not the correct mechanism for managing an internal labeling team.
$19
Get all 235 questions with detailed answers and explanations
Instant download HTML + PDF delivered the moment payment clears.
Secure Stripe checkout we never see or store your card details.
7-day refund if files are defective see our refund policy.