Sample Questions and Answers
Q1. You are designing a data pipeline that ingests streaming data from IoT devices. Which AWS service is best suited for capturing and processing this data in real time?
A. Amazon S3
B. Amazon Kinesis Data Streams
C. AWS Glue
D. Amazon Redshift
Answer: B. Amazon Kinesis Data Streams
Explanation: Kinesis Data Streams is designed for real-time data ingestion and processing, especially suitable for use cases involving IoT telemetry.
Q2. What is the most efficient method to move petabytes of on-premises data to Amazon S3 for a machine learning project?
A. AWS Data Pipeline
B. AWS Snowball
C. Amazon Kinesis
D. AWS DataSync
Answer: B. AWS Snowball
Explanation: AWS Snowball is a physical data transport solution that helps transfer large amounts of data (in TBs or PBs) into AWS efficiently.
Q3. Which format is most efficient for storing large-scale ML datasets in Amazon S3?
A. CSV
B. JSON
C. Parquet
D. TXT
Answer: C. Parquet
Explanation: Apache Parquet is a columnar storage format that provides efficient data compression and encoding, ideal for big data and ML workloads.
Q4. Which AWS service can catalog and search metadata for datasets stored in Amazon S3?
A. AWS DataSync
B. AWS Glue Data Catalog
C. Amazon Athena
D. AWS Lake Formation
Answer: B. AWS Glue Data Catalog
Explanation: The Glue Data Catalog helps store, annotate, and search metadata for datasets across AWS.
Q5. You need to transform and normalize data before training your ML model. Which service would you use for ETL?
A. Amazon QuickSight
B. Amazon SageMaker Processing
C. AWS Glue
D. AWS Lambda
Answer: C. AWS Glue
Explanation: AWS Glue is a serverless ETL service suitable for data cleansing, transformation, and loading tasks for ML workflows.
Q6. Which SageMaker tool allows you to visualize and analyze data within the same environment as your model development?
A. SageMaker Ground Truth
B. SageMaker Studio
C. SageMaker Neo
D. SageMaker Clarify
Answer: B. SageMaker Studio
Explanation: SageMaker Studio provides a web-based interface for end-to-end ML development, including data exploration and visualization.
Q7. During data analysis, you discover high-cardinality categorical variables. Which is a common method to reduce dimensionality?
A. One-hot encoding
B. Min-max normalization
C. Hashing trick
D. Feature scaling
Answer: C. Hashing trick
Explanation: The hashing trick reduces the dimensionality of high-cardinality categorical variables by mapping categories to a fixed-size feature space.
Q8. What is the most appropriate visualization to detect outliers in a numeric feature?
A. Bar chart
B. Line chart
C. Box plot
D. Heatmap
Answer: C. Box plot
Explanation: Box plots effectively show the distribution of data and identify outliers beyond the whiskers.
Q9. You find a feature with zero variance in your dataset. What should you do with it?
A. Scale it
B. Encode it
C. Drop it
D. Impute missing values
Answer: C. Drop it
Explanation: A zero-variance feature contains the same value for all samples and adds no useful information to the model.
Q10. How can you handle missing values in numerical features before model training?
A. Drop entire rows
B. Use mean or median imputation
C. Encode them with -1
D. All of the above
Answer: D. All of the above
Explanation: Depending on the dataset and context, any of these strategies may be appropriate. Imputation is common to preserve data.
Q11. Which algorithm is most suitable for a binary classification problem with highly imbalanced data?
A. Linear Regression
B. Decision Trees
C. XGBoost with class weights
D. K-Means
Answer: C. XGBoost with class weights
Explanation: XGBoost supports handling class imbalance using scale_pos_weight and performs well in such scenarios.
Q12. You want to train a model in SageMaker with automatic hyperparameter tuning. Which feature should you use?
A. SageMaker Clarify
B. SageMaker Debugger
C. SageMaker Automatic Model Tuning
D. SageMaker Neo
Answer: C. SageMaker Automatic Model Tuning
Explanation: Automatic Model Tuning performs hyperparameter optimization using Bayesian search or random search methods.
Q13. What is the primary metric to evaluate a regression model’s performance?
A. F1 Score
B. Precision
C. Mean Squared Error (MSE)
D. ROC AUC
Answer: C. Mean Squared Error (MSE)
Explanation: MSE is commonly used to measure the average squared difference between predicted and actual values in regression tasks.
Q14. Which feature of SageMaker helps monitor loss functions and gradients during model training?
A. SageMaker Experiments
B. SageMaker Clarify
C. SageMaker Debugger
D. SageMaker Model Monitor
Answer: C. SageMaker Debugger
Explanation: SageMaker Debugger allows real-time analysis of model metrics, gradients, and weights during training.
Q15. What is early stopping in machine learning?
A. A method to speed up training
B. A technique to avoid overfitting
C. A way to remove features
D. A feature scaling method
Answer: B. A technique to avoid overfitting
Explanation: Early stopping halts training once the validation loss starts to increase, preventing overfitting.
Q16. Which SageMaker feature enables A/B testing between model variants in production?
A. SageMaker Ground Truth
B. SageMaker Pipelines
C. SageMaker Model Monitor
D. SageMaker Multi-Model Endpoints
Answer: D. SageMaker Multi-Model Endpoints
Explanation: Multi-model endpoints support hosting multiple models and allow versioning and A/B testing.
Q17. Which AWS service automates ML workflows such as preprocessing, training, and deployment?
A. AWS Step Functions
B. SageMaker Pipelines
C. AWS Glue
D. Amazon Kinesis
Answer: B. SageMaker Pipelines
Explanation: SageMaker Pipelines is a CI/CD service specifically designed for ML workflows.
Q18. Which AWS service can be used to host a real-time inference endpoint for a trained model?
A. Amazon Polly
B. Amazon SQS
C. Amazon SageMaker
D. AWS Lambda
Answer: C. Amazon SageMaker
Explanation: SageMaker provides real-time endpoints for deploying and hosting models at scale.
Q19. What does SageMaker Model Monitor primarily track?
A. Deployment uptime
B. Model weights
C. Data drift and prediction bias
D. EC2 usage
Answer: C. Data drift and prediction bias
Explanation: Model Monitor observes model behavior in production and detects drift, bias, and anomalies in data.
Q20. How can you reduce inference costs while serving multiple models on SageMaker?
A. Use multiple endpoints
B. Use SageMaker Ground Truth
C. Use multi-model endpoints
D. Use batch transform jobs
Answer: C. Use multi-model endpoints
Explanation: Multi-model endpoints allow cost-efficient hosting of multiple models under a single endpoint.
Q21. You’re training a deep learning model and want to reduce training time. What SageMaker feature should you consider?
A. SageMaker Ground Truth
B. Elastic Inference
C. Model Monitor
D. SageMaker Debugger
Answer: B. Elastic Inference
Explanation: Elastic Inference attaches inference accelerators to reduce GPU costs and training time.
Q22. Your model is underfitting the data. What is the best first step?
A. Add dropout
B. Reduce training data
C. Increase model complexity
D. Increase regularization
Answer: C. Increase model complexity
Explanation: Underfitting indicates that the model is too simple; increasing its capacity may improve learning.
Q23. A customer wants secure, auditable labeling of medical records. What AWS service is best?
A. Amazon Mechanical Turk
B. SageMaker Ground Truth with private workforce
C. Amazon Comprehend
D. SageMaker Clarify
Answer: B. SageMaker Ground Truth with private workforce
Explanation: For secure, compliant tasks, a private workforce can label sensitive data securely within Ground Truth.
Q24. Which technique can improve generalization in a neural network?
A. Increase learning rate
B. Add L2 regularization
C. Reduce number of layers
D. Train with fewer epochs
Answer: B. Add L2 regularization
Explanation: L2 regularization helps reduce overfitting by penalizing large weights, improving generalization.
Q25. You’re deploying a model that requires latency under 10 ms. Which SageMaker feature should you use?
A. Batch Transform
B. Real-time endpoint
C. Asynchronous inference
D. SageMaker Pipelines
Answer: B. Real-time endpoint
Explanation: Real-time endpoints are designed for low-latency, high-throughput inference.
Q26. What metric is best when evaluating a fraud detection model with a 0.1% positive class?
A. Accuracy
B. Precision
C. F1 Score
D. ROC-AUC
Answer: D. ROC-AUC
Explanation: ROC-AUC evaluates performance across all classification thresholds and is useful for imbalanced datasets.
Q27. Which AWS service allows querying structured data stored in S3 using SQL?
A. Amazon Redshift
B. AWS Glue
C. Amazon Athena
D. AWS Lake Formation
Answer: C. Amazon Athena
Explanation: Athena allows running SQL queries directly on data in S3 without requiring a traditional database.
Q28. What is one benefit of using SageMaker Experiments?
A. Monitoring billing usage
B. Automatically scaling models
C. Tracking training runs and parameters
D. Sharing models with Amazon Marketplace
Answer: C. Tracking training runs and parameters
Explanation: SageMaker Experiments helps track and compare multiple training jobs, improving reproducibility.
Q29. You need to detect and reduce model bias. Which tool is most appropriate?
A. SageMaker Clarify
B. SageMaker Neo
C. SageMaker Debugger
D. SageMaker Pipelines
Answer: A. SageMaker Clarify
Explanation: Clarify provides tools for bias detection and explainability in datasets and models.
Q30. Which AWS service can schedule and orchestrate multiple ML jobs with dependencies?
A. AWS CloudFormation
B. AWS Step Functions
C. Amazon EventBridge
D. Amazon Kinesis
Answer: B. AWS Step Functions
Explanation: Step Functions allow orchestration of multiple AWS services and tasks in serverless workflows, useful for ML job orchestration.
Reviews
There are no reviews yet.