Container Architecture
The Docker container consists of the Assets, scripts to specify its execution and entry points, along with the failsafe mechanisms. This acts as a self sufficient application to take in new data and give out predictions
These assets are necessary files needed to make predictions. the libraries used for preprocessing, scripts that determine how the container app will work, the input drift detection script and obviously the trained ML models.
Assets
Once all the assets are accounted for, the execution script follows the flowchart. This ensures predictions are always of high confidence.
Workflow
When retraining is automatically triggered, a new container image is created with a timestamped tag, enabling easy rollbacks to previous versions if needed.