A proof of concept for implementing MLOps.
Influenced by Made With ML.
- The API model training endpoint could act as a trigger to train a model on a distributed computing network.
- Utility functions should be implemented. This would allow more consistent results across the system when more complicated models are considered.
- Transform function such that data is the same when either training or predicting.
- Add model registry
- Add drift monitoring to dashboard
- Add S3 equivalent for storage between model training phases, model storage etc