odakris/churn-mlops-pipeline

How odakris/churn-mlops-pipeline works

This is a configurable telecom churn-classification application. Its main workflow downloads Telco customer data, cleans and splits it, trains a tuned scikit-learn pipeline with MLflow tracking, and evaluates it; separate serving paths provide JSON/web and batch inference. The graph distinguishes orchestration from the stage implementations and includes configuration, model construction, and persisted artifacts. The README documents several capabilities whose implementation was not sampled; those links are cited to the README rather than presented as inspected source behavior.

Main entry points: config.py Typed configuration · predict.py Batch prediction · train.py Tune and train · evaluate.py Test evaluation

Components 19

Training workflow

Pipeline runner src/run_pipeline.py
Data download src/get_data.py
Clean and split src/preprocess.py
Tune and train src/train.py
Test evaluation src/evaluate.py

Model and configuration

Pipeline builder src/pipeline.py
Typed configuration src/config.py
Project settings configs/config.yaml

Inference interfaces

JSON inference API src/api.py
Web interface src/static/index.html
Batch prediction src/predict.py
Prediction results src/predict.py

Tracking and artifacts

Prepared datasets data
MLflow tracking and registry
Saved model and metadata data
Evaluation reports data

Other components

Project operator
Source dataset
New client CSV

Connections 32

Project operator → Pipeline runner: starts README.md
Pipeline runner → Data download: calls src/run_pipeline.py
Pipeline runner → Clean and split: calls src/run_pipeline.py
Pipeline runner → Tune and train: calls src/run_pipeline.py
Pipeline runner → Test evaluation: calls src/run_pipeline.py
Data download → Source dataset: downloads README.md
Data download → Typed configuration: loads settings src/get_data.py
Clean and split → Typed configuration: loads settings src/preprocess.py
Clean and split → Prepared datasets: writes splits src/preprocess.py
Prepared datasets → Tune and train: provides training data README.md
Prepared datasets → Test evaluation: provides test data README.md
Tune and train → Typed configuration: loads settings src/train.py
Tune and train → Pipeline builder: builds search pipeline src/train.py
Tune and train → Saved model and metadata: writes model metadata src/train.py
Tune and train → MLflow tracking and registry: logs and registers src/train.py
Pipeline builder → Typed configuration: uses model settings src/pipeline.py
Typed configuration → Project settings: loads settings README.md
Test evaluation → Typed configuration: loads settings src/evaluate.py
Test evaluation → Saved model and metadata: loads model src/evaluate.py
Test evaluation → Evaluation reports: writes metrics and plots src/evaluate.py
Test evaluation → MLflow tracking and registry: logs evaluation src/evaluate.py
Project operator → JSON inference API: uses README.md
JSON inference API → Typed configuration: loads settings src/api.py
JSON inference API → Saved model and metadata: loads model src/api.py
JSON inference API → Prepared datasets: reads reference data src/api.py
JSON inference API → Web interface: serves page README.md
Project operator → Batch prediction: runs README.md
Batch prediction → Typed configuration: loads settings src/predict.py
Batch prediction → New client CSV: reads CSV src/predict.py
Batch prediction → Saved model and metadata: loads local model src/predict.py
Batch prediction → MLflow tracking and registry: loads registry model src/predict.py
Batch prediction → Prediction results: writes predictions src/predict.py

Drawn by GitDiagram on November 12, 2026 from the repository's file tree, README and source files. Read as Markdown ·