AWS Certified Machine Learning – SpecialtyMachine Learning Implementation and OperationsEasy
A development team wants to implement a robust MLOps practice where new model versions are automatically retrained and deployed upon detection of data drift or significant performance degradation. This process should be fully automated and trigger subsequent stages only if the previous stage is successful. Which design pattern best describes this automated, event-driven retraining and deployment workflow?
- AManual retraining and deployment
- BStatic model deployment
- CBatch processing with scheduled jobs
- DEvent-driven MLOps pipeline
Show answer & explanationAnswer & explanation
Correct answer: D. Event-driven MLOps pipeline
An event-driven MLOps pipeline uses events (like data drift detection or performance degradation) to automatically trigger subsequent stages of the ML lifecycle, such as retraining, evaluation, and deployment, ensuring continuous improvement.
Why the other options are wrong
- A. Manual processes contradict the requirement for 'fully automated'.
- B. Static model deployment implies no automated retraining or updates, which is contrary to the scenario.
- C. Batch processing with scheduled jobs is not 'event-driven' and may not react immediately to drift.
Event-Driven MLOps Pipeline
An MLOps architecture where machine learning lifecycle stages (e.g., retraining, deployment) are automatically triggered by specific events.
- Automates reactions to data drift, performance degradation, or new data availability.
- Ensures models are continuously updated and relevant.
- Often uses services like AWS Lambda, EventBridge, and SageMaker Model Monitor.
Memory trick: Events ring the bell for automated ML actions.