AWS ML Model Deployment and Inference Quiz

1. Which metric is most important for monitoring real-time endpoint performance?

Model invocation count and latency

Training time

Data upload speed

Hyperparameter values

2. Which SageMaker feature automatically detects model performance degradation over time?

Model Monitor Baseline

CloudWatch Alarms

Lambda Functions

Step Functions

3. What is the role of Amazon S3 in SageMaker inference workflows?

Stores model artifacts, input data for batch transform, and predictions

Provides real-time inference only

Trains models directly

Monitors endpoint performance

4. For serverless inference on AWS, which service should you use?

SageMaker Serverless Endpoints

SageMaker Realtime Endpoints

EC2 On-Demand

Batch Transform only

5. Which optimization technique reduces model size by removing less important weights?

Pruning

Quantization

Distillation

Normalization

6. What does the SageMaker Model Registry provide?

Version control and metadata management for trained models

Automatic model training

Real-time data labeling

Endpoint monitoring dashboards

7. In production, which approach ensures minimal downtime when updating a model endpoint?

Blue-green deployment with traffic shifting

Immediately replacing the old model

Stopping all traffic before deployment

Using batch transform only

8. What is the advantage of using SageMaker Feature Store?

Provides centralized feature management for training and inference

Automatically trains models

Reduces storage costs

Eliminates data preprocessing

9. Which service provides real-time monitoring of deployed models for data drift?

SageMaker Model Monitor

CloudWatch

AWS Config

VPC Flow Logs

10. What is the purpose of an inference pipeline in SageMaker?

Chain preprocessing, inference, and postprocessing in a single endpoint

Store training data

Monitor model drift

Schedule batch jobs

11. Which AWS service is primarily used to deploy machine learning models for real-time inference?

Amazon SageMaker

AWS Lambda

Amazon EC2

AWS Glue

12. In SageMaker, what is A/B testing used for?

Comparing model performance by routing traffic to multiple variants

Testing data preprocessing steps

Validating model artifacts

Monitoring endpoint health

13. What does model quantization achieve?

Reduces model size and inference latency with minimal accuracy loss

Increases model accuracy

Improves training speed

Adds more layers to the model

14. Which AWS service can invoke a SageMaker endpoint for inference?

Lambda

API Gateway

Step Functions

All of the above

15. What is batch transform used for in SageMaker?

Making predictions on large datasets without needing an endpoint

Real-time single prediction requests

Training models on batches

Monitoring endpoint performance

16. Which container image format does SageMaker use for custom inference code?

Docker

JAR

AMI

Lambda function

17. In SageMaker, what is a model artifact?

The serialized model file after training

The training dataset

The inference configuration

The CloudWatch logs

18. What does SageMaker Neo do?

Compiles models to run on edge devices and reduces inference latency

Automatically trains ML models without code

Manages data labeling workflows

Monitors model performance in production

19. Which feature allows you to automatically scale SageMaker endpoints based on traffic?

Target tracking scaling

Manual scaling

Batch transformation

Data pipeline scaling

20. What is the primary purpose of SageMaker Multi-Model Endpoints?

Host multiple models on a single endpoint to reduce costs

Train multiple models in parallel

Deploy models across multiple regions

Increase inference latency for accuracy

AWS ML Engineer Model Deployment and Inference Quiz