A retail company is creating an AI-powered assistant for customers. The company has a large body of documentation that the assistant needs to use for general inquiries. The company wants any responses about prices to use only documentation that is less than 1 month old.
Which solution will meet these requirements?
A. Use Amazon Q Business to develop the responses. Configure a document attribute filter so that responses about prices use only the documents from the past month.
B. Use Amazon Q Business to develop the responses. Configure the source attribution citation so that responses about prices use only the documents from the past month.
C. Segment the documents into folders based on the month of document creation. Configure Amazon Q Developer to use only the documents from the past month to develop responses about prices.
D. Segment the documents into folders based on the month of document creation. Grant the assistant access to only the documents from the past month for responses about prices. Use Amazon Q Developer to develop the responses.
Show Answer
Correct Answer: A
Explanation: Amazon Q Business is designed for enterprise document-based question answering and supports document attribute (metadata) filtering to restrict retrieval based on attributes such as document dates. This allows price-related responses to be grounded only in documents less than one month old. Source attribution only cites retrieved documents and does not control retrieval, and Amazon Q Developer is intended for developer productivity rather than enterprise knowledge assistants.
Question 32
A healthcare company uses an Amazon SageMaker AI endpoint to host a model that predicts patient readmission risk to hospitals. The company wants to predict patient readmissions with high accuracy and is willing to tolerate false positives. The current model performance has degraded over the previous year.
The company trains and deploys a new model as a shadow variant for testing on live traffic from hospitals. The company monitors the performance of the new model for a month. During the month of testing, the shadow variant has a higher recall than the existing model but has a lower precision.
What should the company do next?
A. Promote the shadow variant to full production.
B. Extend the shadow testing period to capture more data. Monitor the new model to determine whether precision improves.
C. Use a blue/green deployment strategy to allocate a small percentage of traffic to the shadow variant to reduce model errors.
D. Disable the shadow variant and roll back to the main variant.
Show Answer
Correct Answer: A
Explanation: The stated business preference is to maximize detection of patient readmissions and tolerate false positives. Higher recall means the new model identifies more actual readmissions, while lower precision means more false positives, which the company has indicated it can accept. After a month of successful shadow testing on live traffic, promoting the shadow variant is the appropriate next step.
Question 33
A company has built, trained, and tuned two new ML models:
• Model A detects if a transaction is fraudulent based on the IP address, location, and user credentials. This model will be accessed every time a transaction occurs.
• Model B forecasts sales totals for the next month based on historical sales data. This model will be accessed one time each month.
The company must deploy both models to production by using Amazon SageMaker AI.
Which hosting solution for the models should the company use to meet these requirements?
A. Host both models in one container behind one real-time endpoint.
B. Host Model A with an asynchronous endpoint. Host Model B with a real-time endpoint.
C. Host Model A with a real-time endpoint. Use batch transform for Model B.
D. Use batch transform for Model A. Host Model B with an asynchronous endpoint.
Show Answer
Correct Answer: C
Explanation: Model A performs online fraud detection for every transaction and requires low-latency synchronous inference, making a real-time endpoint the appropriate choice. Model B runs only once per month on historical data, so Batch Transform is more cost-effective because it performs offline inference without maintaining a persistent endpoint.
Question 34
A company stores user clickstream data in an Amazon S3 bucket in AWS Account A. The company needs to use the data to train an ML model in Amazon SageMaker AI in AWS Account B. The training will take 10 days.
The company needs to use only private IP addresses in the training. The company also must make sure that no training metadata is shared with AWS.
Which solution will meet these requirements?
A. Set up VPC peering between Account A and Account B. Contact AWS by email to opt out of metadata collection.
B. Set up a VPC endpoint for the S3 bucket. Set the SageMaker AI OPT_OUT_TRACKING environment variable to 1 in the training job.
C. Configure a security group policy that is assigned to the S3 bucket in Account A to allow access from only Account B. Create AI services opt-out policies.
D. Generate presigned URLs with expiration times for the objects that are stored in the S3 bucket. Access the data by using the presigned URLs. Set the SageMaker AI OPT_OUT_TRACKING environment variable to 1 in the training job.
Show Answer
Correct Answer: B
Explanation: An S3 VPC endpoint (gateway endpoint) allows Amazon SageMaker training jobs running in a VPC to access Amazon S3 over the AWS private network rather than the public internet, satisfying the private IP requirement. Setting the SageMaker OPT_OUT_TRACKING environment variable to 1 disables sharing of training metadata with AWS. The other options rely on unsupported or incorrect mechanisms: VPC peering alone does not provide private S3 access, S3 buckets do not use security groups, and presigned URLs do not meet the private networking requirement.
Question 35
An airline company uses an ML model to adjust ticket prices based on demand. The model runs on Amazon SageMaker real-time endpoints. During previous deployments, the model failed to scale quickly enough when website traffic increased, which caused delays in price adjustments.
An ML engineer needs to configure auto scaling for the SageMaker endpoints to respond rapidly to traffic changes. The solution must use target tracking scaling policies.
Which configuration will be MOST responsive to sudden changes in traffic?
A. Configure auto scaling based on the SageMaker AI InvocationsPerInstance standard metric. Configure 10-second interval resolution, and set the default 300-second scale-in cooldown period.
B. Configure auto scaling based on the SageMaker AI InvocationsPerInstance metric. Configure high-resolution 10-second intervals, and set a 600-second scale-in cooldown period.
C. Configure auto scaling based on the SageMaker InvocationsPerInstance standard metric. Configure 10-second intervals resolution, and set a 600-second scale-in cooldown period.
D. Configure auto scaling based on the SageMaker InvocationsPerInstance metric. Configure high-resolution 10-second intervals, and set the default 300-second scale-in cooldown period.
Show Answer
Correct Answer: D
Explanation: For SageMaker endpoint target tracking, the appropriate predefined scaling metric is InvocationsPerInstance. To maximize responsiveness, use high-resolution (10-second) metrics so scaling decisions are made more frequently. Between the cooldown options, the shorter default 300-second scale-in cooldown is more responsive than 600 seconds while still allowing rapid scale-out behavior. The options referring to 'SageMaker AI InvocationsPerInstance' do not match the standard SageMaker endpoint scaling metric.
Question 36
A company has trained an ML model that is packaged in a container. The company will integrate the model with an existing Python web application. The company needs to host the model on AWS by using Kubernetes.
The company does not want to manage the control plane and must provision the resources in a repeatable manner. The infrastructure must be provisioned by using Python.
Which solution will meet these requirements?
A. Use AWS CloudFormation to provision Amazon EC2 instances in multiple Availability Zones. Set up a Kubernetes cluster. Host the model container on the Kubernetes cluster.
B. Use the AWS CLI to provision an Amazon Elastic Kubernetes Service (Amazon EKS) cluster. Store the image in an Amazon Elastic Container Registry (Amazon ECR) repository. Host the model container on the EKS cluster.
C. Use the AWS Cloud Development Kit (AWS CDK) to provision an Amazon Elastic Kubernetes Service (Amazon EKS) cluster. Store the image in an Amazon Elastic Container Registry (Amazon ECR) repository. Host the model container on the EKS cluster.
D. Use AWS CloudFormation to provision an Amazon Elastic Kubernetes Service (Amazon EKS) cluster. Store the image in an Amazon Elastic Container Registry (Amazon ECR) repository. Host the model container on the EKS cluster.
Show Answer
Correct Answer: C
Explanation: Amazon EKS provides a managed Kubernetes control plane, satisfying the requirement to avoid managing the control plane. The infrastructure must be provisioned in a repeatable manner using Python, which is a core capability of AWS CDK. Storing the container image in Amazon ECR and deploying it to EKS is the standard AWS container workflow.
Question 37
A company is developing an internal cost-estimation tool that uses an ML model in Amazon SageMaker AI. Users upload high-resolution images to the tool.
The model must process each image and predict the cost of the object in the image. The model also must notify the user when processing is complete.
Which solution will meet these requirements?
A. Store the images in an Amazon S3 bucket. Deploy the model on SageMaker AI. Use batch transform jobs for model inference. Use an Amazon Simple Queue Service (Amazon SQS) queue to notify users.
B. Store the images in an Amazon S3 bucket. Deploy the model on SageMaker AI. Use an asynchronous inference strategy for model inference. Use an Amazon Simple Notification Service (Amazon SNS) topic to notify users.
C. Store the images in an Amazon Elastic File System (Amazon EFS) file system. Deploy the model on SageMaker AI. Use batch transform jobs for model inference. Use an Amazon Simple Queue Service (Amazon SQS) queue to notify users.
D. Store the images in an Amazon Elastic File System (Amazon EFS) file system. Deploy the model on SageMaker AI. Use an asynchronous inference strategy for model inference. Use an Amazon Simple Notification Service (Amazon SNS) topic to notify users.
Show Answer
Correct Answer: B
Explanation: Amazon SageMaker Asynchronous Inference is designed for requests with large payloads (such as high-resolution images) and long processing times. Input data is commonly stored in Amazon S3, and asynchronous inference writes results to S3 while supporting completion notifications through Amazon SNS. Batch Transform is intended for offline batch processing rather than user-driven request/notification workflows, and EFS is not the standard storage choice for asynchronous inference inputs.
Question 38
An ML engineer is configuring auto scaling for an inference component of a model that runs behind an Amazon SageMaker AI endpoint. The ML engineer configures SageMaker AI auto scaling with a target tracking scaling policy set to 100 invocations per model per minute. The SageMaker AI endpoint scales appropriately during normal business hours. However, the ML engineer notices that at the start of each business day, there are zero instances available to handle requests, which causes delays in processing.
The ML engineer must ensure that the SageMaker AI endpoint can handle incoming requests at the start of each business day.
Which solution will meet this requirement?
A. Reduce the SageMaker AI auto scaling cooldown period to the minimum supported value. Add an auto scaling lifecycle hook to scale the SageMaker AI instances.
B. Change the target metric to CPU utilization.
C. Modify the scaling policy target value to one.
D. Apply a step scaling policy that scales based on an Amazon CloudWatch alarm. Apply a second CloudWatch alarm and scaling policy to scale the minimum number of instances from zero to one at the start of each business day.
Show Answer
Correct Answer: D
Explanation: Target tracking can scale an endpoint down to zero, leaving no warm capacity at the beginning of the day. Reducing cooldowns or changing the target metric/value does not proactively create capacity before traffic arrives. A scheduled CloudWatch alarm with a step scaling policy can increase the minimum capacity from 0 to 1 before business hours, while another scaling policy continues to handle demand-driven scaling.
Question 39
HOTSPOT
-
An ML engineer needs to use Amazon SageMaker hyperparameter tuning to reduce the training time for an ML model.
Select and order the correct steps from the following list to meet this requirement. Each step should be selected one time or not at all. (Select and order three.)
• Choose Bayesian optimization and increase the number of parameters.
• Choose Hyperband tuning and decrease the number of parameters.
• Choose random search and use a random seed of -1.
• Deploy the model to a SageMaker endpoint.
• Evaluate the change in training time.
• Retrain the model.
Show Answer
Correct Answer: Step 1: Choose Hyperband tuning and decrease the number of parameters.
Step 2: Retrain the model.
Step 3: Evaluate the change in training time.
Explanation: Hyperband reduces training time through early stopping of poor-performing trials. Reducing the hyperparameter search space further decreases tuning time. After applying the new tuning strategy, retrain and then evaluate whether training time has improved.
Question 40
An ML engineer wants to use Amazon SageMaker Data Wrangler to perform preprocessing on a dataset. The ML engineer wants to use the processed dataset to train a classification model. During preprocessing, the ML engineer notices that a text feature has a range of thousands of values that differ only by spelling errors. The ML engineer needs to apply an encoding method so that after preprocessing is complete, the text feature can be used to train the model.
Which solution will meet these requirements?
A. Perform ordinal encoding to represent categories of the feature.
B. Perform similarity encoding to represent categories of the feature.
C. Perform one-hot encoding to represent categories of the feature.
D. Perform target encoding to represent categories of the feature.
Show Answer
Correct Answer: B
Explanation: Similarity encoding is designed for high-cardinality categorical/text features where many values are similar due to spelling variations or typos. It encodes categories based on string similarity, allowing near-duplicate values to have similar representations. Ordinal, one-hot, and target encoding do not address spelling-error similarity and are less suitable for this use case.
$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.
