How vikashishere/yt-capstone-project works
Project type and purpose
• This is an end-to-end MLOps capstone project that encompasses data ingestion, preprocessing, feature engineering, model training, evaluation, registration, and deployment.
• It includes a model serving layer (a Flask application), versioned pipelines (DVC + Makefile), CI (GitHub Actions), containerization (Dockerfile), and deployment manifest (deployment.yaml).
File structure
• Top-level orchestration:
– dvc.yaml, dvc.lock, .dvc/ : defines and tracks the data pipeline stages
– Makefile / docs/Makefile : build and docs targets
– Dockerfile / deployment.yaml : container build and k8s deployment
– .github/workflows_stop/ci.yaml : continuous integration pipeline
• Source code:
– src/ :
· connections/ : database or cloud storage connectors (S3, SSMS)
· data/ : ingestion and preprocessing routines
· features/ : feature engineering
· model/ : train_model, model_building, evaluation, register_model, predict_model
· visualization/ : result plotting
· logger/ : logging setup
• Model serving:
– flask_app/ : Flask API endpoints, HTML template
• Experiments & reports:
– notebooks/ : exploratory experiments (IMDB sentiment data)
– reports/ : figures and write-ups
• Utilities:
– scripts/promote_model.py : script to move a model from staging to production
– tests/ & test_environment.py : unit/integration tests
– params.yaml, requirements.txt, setup.py : config, dependencies, packaging
README
• Minimal; confirms “end to end mlops capstone project for educational purpose.”
• Implies demonstration of MLOps best practices, pipeline orchestration, model serving and deployment.