AWS Certified Machine Learning – SpecialtyMachine Learning Implementation and OperationsMedium
A data engineering team is setting up an MLOps pipeline for a fraud detection model. They need to ensure that feature data used for training and inference is consistent, versioned, and easily accessible across different stages of the ML lifecycle. The features should also be available for both batch processing (training) and low-latency real-time lookups (inference). Which AWS service is best suited for this requirement?
- AAmazon SageMaker Feature Store
- BAmazon S3
- CAmazon DynamoDB
- DAmazon Redshift
Show answer & explanationAnswer & explanation
Correct answer: A. Amazon SageMaker Feature Store
Amazon SageMaker Feature Store is specifically designed to address the challenges of feature management for ML. It provides a centralized repository for features, ensuring consistency between training and inference, enabling versioning, and offering both an offline store for batch processing and an online store for low-latency real-time lookups.
Why the other options are wrong
- B. S3 is an object storage service and can store feature data, but it does not provide the online/offline store, consistency guarantees, or specialized API for feature management that Feature Store does.
- C. DynamoDB can provide low-latency access but lacks the built-in feature engineering, versioning, and offline/online store capabilities tailored for ML that Feature Store offers.
- D. Redshift is a data warehouse suitable for large-scale analytics and batch processing, but not optimized for low-latency real-time feature lookups or specific ML feature management.
SageMaker Feature Store
A fully managed service that provides a centralized repository for machine learning features, enabling consistent, versioned, and low-latency access for both training and inference.
- Ensures consistency between training and inference features.
- Provides an online store for low-latency inference and an offline store for training.
- Supports feature versioning and discovery.
Memory trick: Feature Store: the brain's pantry for ML models.