How muhammad-ahmad2511/telco-churn-mlops works
This repository is presented as an end-to-end Telco churn MLOps product: data preparation and feature materialization lead conceptually to model development, deployment for inference, and operational monitoring. The tree identifies those stages and the README names DVC, MLflow, Airflow, Kubeflow, Flask, BentoML, Docker, Kubernetes, AWS, and Prometheus/Grafana. Source excerpts are empty, so internal calls and concrete orchestration links cannot be verified; dashed, un evidenced edges indicate only the broad documented workflow.
Main entry points: train.py Model training · app.py Flask API · ingestion.py Data ingestion · validation.py Data validation
Components 20
Data and features
- Data ingestion
src/data/ingestion.py - Data validation
src/data/validation.py - Preprocessing
src/features/preprocessing.py - Feature materialization
feature_store/materialize.py
Model development
- Hyperparameter tuning
src/models/tune.py - Model training
src/models/train.py - Model evaluation
src/models/evaluate.py
Inference
- Flask API
src/serving/app.py - BentoML service
src/serving/bentoml_service.py - Model loading
src/serving/model_loader.py
MLOps operations
- MLflow
- DVC pipeline
dvc.yaml - Airflow orchestration
pipelines/airflow - Kubeflow pipelines
pipelines/kubeflow - Monitoring
src/monitoring/monitor.py - Prometheus/Grafana
- Docker runtime
docker/docker-compose.yml - Kubernetes deployment
k8s - AWS
Other components
- Prediction client
Connections 16
- Prediction client → Flask API: requests predictions
- Data ingestion → Data validation: validate data
- Data validation → Preprocessing: prepare features
- Preprocessing → Feature materialization: materialize features
- Model training → Hyperparameter tuning: tune model
- Model training → Model evaluation: evaluate model
- Model training → MLflow: track experiments
- DVC pipeline → Data ingestion: orchestrates pipeline
- Airflow orchestration → Model training: orchestrates
- Kubeflow pipelines → Model training: orchestrates
- Flask API → Model loading: loads model
- BentoML service → Model loading: loads model
- Monitoring → Prometheus/Grafana: exports metrics
- Docker runtime → Flask API: packages service
- Kubernetes deployment → Docker runtime: deploys runtime
- AWS → Kubernetes deployment: hosts deployment