wayground logo

Free Printable Worksheets

NEW

Font size

S
M
L
XL
Worksheets

MLOps Workshop Assessment – AWS SageMaker MLOps Pipeline

Total questions: 29

Worksheet time: 15mins

Name
Class
Date
1.

Architecture & Pipeline Design: Your medical AI team needs to implement an automated retraining pipeline that triggers when new hospital data arrives. Which AWS services combination provides the most robust event-driven MLOps architecture?

a)

CloudWatch Events → Lambda → SageMaker Training → Manual Model Registry

b)

S3 Event Notifications → Lambda → SageMaker Pipeline → Model Registry → Conditional Deployment

c)

API Gateway → EC2 → Manual Training → S3 Storage

d)

SageMaker Studio → Manual Triggers → Direct Endpoint Deployment

2.

Model Performance & Quality Gates: Your cancer detection model achieves 94% accuracy, but your quality gate is set to 95%. What should happen in an automated MLOps pipeline?

a)

Deploy the model since 94% is close to 95%

b)

The pipeline should automatically reject the model

c)

Manually override the quality gate for this deployment

d)

Deploy only to development environment

3.

Training Job Troubleshooting: Your SageMaker training job fails with "AlgorithmError: ExecuteUserScriptError: ExitCode 2". The CloudWatch logs show "ModuleNotFoundError: No module named 'tensorflow'". What is the most likely root cause?

a)

Insufficient IAM permissions

b)

Wrong instance type selected

c)

Training script dependencies not properly configured

d)

S3 bucket permissions issue

4.

Medical AI Compliance & Governance: For healthcare applications, which practice is MOST critical for regulatory compliance?

a)

High accuracy metrics and fast inference times

b)

Model interpretability and audit logs

c)

Cost optimization and resource scaling

d)

Automated deployment without human oversight

5.

Model Registry & Versioning: You have deployed cancer-detection-model-v1.2 to production. A new model version v1.3 achieves higher accuracy but shows concerning behavior on some test cases. What's the best MLOps strategy?

a)

Immediately replace v1.2 with v1.3 due to higher accuracy

b)

Keep v1.2 in production and investigate v1.3 before deployment

c)

Deploy both models and let users choose

d)

Deploy v1.3 immediately to get the accuracy benefits

6.

Data Pipeline & Storage: Your medical imaging data is stored in S3 with the structure: "s3://bucket/data/train/benign/" and "s3://bucket/data/train/malignant/". What's the recommended approach for SageMaker training?

a)

Download all images to local storage before training

b)

Use S3 as direct input with SageMaker's built-in data loading

c)

Convert images to a different format first

d)

Stream images through API Gateway during training

7.

Endpoint Configuration & Scaling: Your cancer detection endpoint needs to handle varying traffic throughout the day. What's the most appropriate scaling approach?

a)

Use a fixed large instance that can handle peak traffic

b)

Use auto-scaling to adjust capacity based on demand

c)

Manually scale up and down based on time of day

d)

Use the smallest instance to minimize costs

8.

MLOps Monitoring: Which metric is MOST important to monitor for a production medical AI model?

a)

Cost per prediction

b)

Model accuracy over time

c)

Server CPU utilization

d)

Number of API calls per day

9.

Security & IAM Troubleshooting: Your Lambda function fails to trigger SageMaker training with error: "User is not authorized to perform: sagemaker:CreateTrainingJob". What's the most likely issue?

a)

The Lambda needs SageMaker permissions in its execution role

b)

S3 bucket policy needs to allow Lambda access

c)

SageMaker service is down

d)

The training data is corrupted

10.

Model Evaluation & Testing: During model evaluation, you discover your cancer detection model performs differently across different hospitals. What should you do first?

a)

Deploy the model anyway since overall accuracy is good

b)

Investigate the data differences between hospitals

c)

Train separate models for each hospital

d)

Ignore the differences and focus on overall performance

11.

Pipeline Orchestration: Your MLOps pipeline includes: Data Validation → Training → Evaluation → Deployment. The evaluation step fails due to missing test data. How should the pipeline handle this?

a)

Skip evaluation and proceed to deployment

b)

Stop the pipeline and alert the team

c)

Use training data for evaluation instead

d)

Deploy without evaluation this time

12.

Cost Optimization: Your training jobs are taking longer than expected and costing more than budgeted. What's the first thing you should investigate?

a)

Switch to cheaper instance types

b)

Reduce the dataset size

c)

Check if the training script is efficient

d)

Train less frequently

13.

Real-time Inference: Your cancer detection API needs to respond quickly for real-time diagnosis support. The current response time is too slow. What should you try first?

a)

Use a faster instance type for the endpoint

b)

Reduce the model size

c)

Cache previous predictions

d)

Use batch processing instead

14.

Model Rollback: Your production cancer detection model starts showing problems after deployment. You need to rollback to the previous version quickly. What's the fastest approach?

a)

Retrain the previous model version

b)

Update the endpoint to use the previous model

c)

Delete everything and start over

d)

Wait for the issues to resolve themselves

15.

MLOps Best Practices: You're setting up a complete MLOps solution for medical image classification. Which component is most essential for getting started?

a)

Advanced experiment tracking with multiple frameworks

b)

Automated model deployment with proper testing

c)

Complex multi-region disaster recovery

d)

Integration with different hospital systems

16.

Model Performance & Quality Gates: In a medical AI pipeline with safety thresholds, what should happen when a model fails to meet the defined threshold to prevent deployment of substandard models?

a)

Automatically reject the model

b)

Deploy the model with enhanced monitoring

c)

Proceed to manual review before deployment

d)

Delay deployment but keep the model in production

17.

Training Job Troubleshooting: A TensorFlow training job fails with an import error indicating missing packages. What is the most likely root cause to address first?

a)

Insufficient GPU memory on the instance

b)

Training script dependencies not properly configured

c)

Incorrect S3 bucket permissions

d)

Model hyperparameters causing overfitting

18.

Medical AI Compliance & Governance: Which capability is most critical to satisfy healthcare regulations beyond performance metrics?

a)

Batch inference throughput reports

b)

Model interpretability and audit logs

c)

Continuous hyperparameter tuning

d)

Multi-region endpoint failover

19.

Model Registry & Versioning: A medical AI team sees concerning behavior in candidate version v1.3 while v1.2 is stable in production. What should they do before deploying v1.3?

a)

Promote v1.3 immediately to production

b)

Keep v1.2 in production and investigate v1.3 before deployment

c)

Retire v1.2 and run only v1.3 in staging

d)

Rollback v1.2 to v1.1 for safety

20.

Data Pipeline & Storage: For training data access in SageMaker, what is the recommended approach to optimize loading and caching?

a)

Use a self-managed NFS mounted on the training containers

b)

Use S3 as direct input with SageMaker’s built-in data loading

c)

Copy data to EBS volumes pre-training

d)

Stream data from a relational database via JDBC

21.

Endpoint Configuration & Scaling: To improve cost efficiency and performance based on actual demand patterns, what scaling approach should be used for the endpoint?

a)

Provision a fixed number of instances based on peak load

b)

Use auto-scaling to adjust capacity based on demand

c)

Manually scale instances daily

d)

Disable scaling and rely on caching

22.

MLOps Monitoring: In medical applications, which metric is most critical to monitor in production ML systems to detect model drift and degradation?

a)

CPU utilization of endpoints

b)

Model accuracy over time

c)

Number of API requests per minute

d)

Average inference latency only

23.

Security & IAM Troubleshooting: A Lambda function needs to create SageMaker training jobs but fails with AccessDenied. What is the most likely permission issue?

a)

The Lambda needs SageMaker permissions in its execution role

b)

The Lambda must run in a VPC

c)

The training script lacks S3 read access

d)

The endpoint does not allow public access

24.

Model Evaluation & Testing: Performance differs across hospitals during validation. What is the priority action before deployment?

a)

Deploy and monitor in production

b)

Investigate the data differences between hospitals

c)

Increase training epochs to improve accuracy

d)

Reduce the validation dataset to speed evaluation

25.

Pipeline Orchestration: The evaluation step cannot find required data. What should an MLOps pipeline do to ensure safe operation?

a)

Skip the evaluation step and continue

b)

Stop the pipeline and alert the team

c)

Retry silently without notification

d)

Proceed to deploy with the last successful model

26.

Cost Optimization: Training is slow and expensive without clear infrastructure bottlenecks. What should be checked first?

a)

Move training to larger GPU instances

b)

Switch storage from S3 to EFS

c)

Check if the training script is efficient

d)

Increase batch size dramatically

27.

Real-time Inference: Immediate latency improvements are needed for an endpoint. What is the most straightforward action?

a)

Use a faster instance type for the endpoint

b)

Retrain the model with quantization

c)

Enable multi-AZ deployment

d)

Add asynchronous inference

28.

Model Rollback: A newly deployed model underperforms. What is the fastest way to restore the previous stable version?

a)

Recreate all endpoint infrastructure from scratch

b)

Update the endpoint to use the previous model

c)

Roll back the training dataset

d)

Disable auto-scaling temporarily

29.

MLOps Best Practices: For reliable releases, which foundational capability should be implemented first?

a)

Advanced feature engineering pipeline

b)

Automated model deployment with proper testing

c)

Canary traffic shifting across regions

d)

Real-time drift remediation