MLOps with AWS

Which SageMaker extension allows Kubernetes-based ML platforms to integrate with SageMaker services?
A) SageMaker Model Monitor
B) SageMaker Operators
C) SageMaker Pipelines SDK
D) Amazon EKS Optimizer

Where are Docker images typically stored before being pulled to run on AWS services like ECS or SageMaker?
A) Amazon Redshift
B) Amazon Elastic Container Registry (ECR)
C) Amazon DynamoDB
D) AWS Step Functions

Which benefit does Docker provide compared to traditional virtual machines (VMs)?
A) Full guest operating system isolation
B) Reduced resource usage by sharing the host OS
C) Automatic deployment to S3 storage
D) Native encryption of container layers

Which AWS service allows you to run containers without managing servers or clusters?
A) Amazon EKS
B) Amazon ECS with Fargate
C) Amazon SageMaker Studio
D) AWS Lambda

Which AWS service provides a fully managed Kubernetes control plane?
A) Amazon EKS
B) Amazon ECS
C) Amazon SageMaker Pipelines
D) AWS Elastic Beanstalk

Which component in Kubernetes is responsible for ensuring a containerized application runs as intended across a cluster?
A) Service
B) Pod
C) Node
D) Deployment

What is a key advantage of using Fargate with ECS?
A) Provides SSH access to underlying EC2 instances
B) Manages compute resources automatically for containers
C) Allows GPU acceleration for training models
D) Requires manual provisioning of EC2 worker nodes

In a typical MLOps architecture, which service is primarily responsible for managing feature storage and access during model training and inference?
A) Amazon CloudFront
B) Amazon EFS
C) Amazon Feature Store
D) Amazon Aurora

What is the primary purpose of Amazon SageMaker Pipelines?
A) To create model training dashboards
B) To automate and manage ML workflows including preprocessing, training, and deployment
C) To replace Amazon SageMaker Studio for development
D) To serve containerized applications at scale

Which TWO of the following are valid components (steps) in a SageMaker Pipeline? (Select TWO)
A) Processing Step
B) Deployment Step
C) CloudFormation Step
D) Tuning Step
E) Billing Step

In SageMaker Pipelines, what does a Condition Step allow you to do?
A) Merge multiple pipelines into one
B) Encrypt pipeline outputs using KMS
C) Branch the pipeline execution flow based on evaluation metrics or parameters
D) Schedule pipeline runs at specific intervals

Which AWS services commonly integrate with SageMaker Pipelines for CI/CD in MLOps workflows? (Select TWO)
A) AWS CodePipeline
B) AWS CloudWatch
C) AWS CodeBuild
D) AWS Direct Connect
E) Amazon Textract

What is the primary purpose of the SageMaker Model Registry in MLOps workflows?
A) To automatically retrain models on new data
B) To store, version, and track metadata of trained models
C) To monitor endpoint latency and error rates
D) To deploy containerized models at scale

Which TWO features does the SageMaker Model Registry provide? (Select TWO)
A) Model versioning and metadata tracking
B) Automated hyperparameter tuning
C) Model lineage tracking
D) Integrated feature engineering

What role does Amazon SageMaker Feature Store play in an MLOps pipeline?
A) Encrypting model artifacts for secure storage
B) Storing and managing features for training and inference
C) Deploying and scaling containerized endpoints
D) Orchestrating CI/CD pipelines for ML models

What is a primary benefit of implementing automated deployment triggers in an MLOps workflow?
A) Reducing manual intervention and ensuring consistent, repeatable deployments
B) Lowering the cost of model training
C) Increasing infrastructure complexity for greater control
D) Guaranteeing zero-downtime updates without the need for testing

How does SageMaker Model Monitor help detect model drift?
A) By encrypting training data before use
B) By continuously logging and comparing inference input/output metrics against a baseline
C) By automatically retraining the model in production
D) By managing container scaling based on CPU utilization

Which TWO AWS services are commonly integrated with SageMaker for comprehensive monitoring and auditing of ML workflows? (Select TWO)
A) Amazon CloudWatch
B) Amazon CloudTrail
C) Amazon S3 Glacier
D) AWS CodeDeploy

A company with an existing Kubernetes-based ML platform wants to leverage AWS’s managed services to scale its ML workflows without replacing their current infrastructure. Which solution best enables this hybrid approach?
A) Rebuilding the entire system on Amazon SageMaker Studio
B) Using SageMaker Operators for Kubernetes to integrate with existing Kubeflow pipelines
C) Migrating all workloads to AWS Fargate exclusively
D) Deploying Docker containers directly on Amazon EC2 instances without Kubernetes

A development team wants to manage both their infrastructure and application code using a familiar programming language and high-level constructs while deploying an ML application. Which AWS service best meets this requirement?
A) AWS Serverless Application Model (SAM)
B) AWS CloudFormation directly via JSON/YAML templates
C) AWS Cloud Development Kit (CDK)
D) AWS CodePipeline exclusively

Your organization needs to orchestrate a complex ML training workflow that includes data processing, model training, and post-processing with robust error handling and retry mechanisms. Which AWS service should you use to design this workflow?
A) Amazon EventBridge
B) AWS Step Functions
C) AWS Batch
D) Amazon CloudWatch