Customer churn API .
Beyond the notebook, into an API.
Concept illustration / A view of the idea, not a product screenshot.
01 / The context
The problem
worth working on.
A trained model is not yet something a product can use. The learning goal was to connect experimentation to an interface that can accept data and return predictions consistently.
02 / The approach
How the pieces
come together.
- 01
Worked through customer-data preparation, model training, and evaluation as an end-to-end ML exercise.
- 02
Preserved preprocessing and model artifacts for reuse during inference.
- 03
Wrapped predictions in a FastAPI service and containerized the application.
03 / A key decision
The detail
that shapes the system.
Treat preprocessing as part of the model contract. Training and inference need the same transformation rules.
04 / The result
What came out of it.
A first complete ML service that connected model development to application engineering.
Scope & limitations
A learning project, not evidence of a production retention system. Production monitoring and model-drift detection were outside its scope. No live demo is advertised.