Detailed Introduction
MLOps Basics is an open-source learning repository by Raviraja that guides engineers through practical, week-by-week exercises to build production-ready machine learning workflows. The project is organized by weekly modules covering data ingestion and preprocessing, model training, monitoring (Weights & Biases), configuration management (Hydra), data versioning (DVC), model packaging (ONNX, Docker), CI/CD (GitHub Actions), and production monitoring (Kibana), enabling reproducible engineering practices.
Main Features
- Structured weekly tutorials that progressively cover core MLOps stages with runnable examples.
- Practical sample code and environment configurations including DVC, Docker, and CI pipelines.
- Focus on reproducibility and engineering, demonstrating model saving, export, and end-to-end deployment.
- Open-source license that allows community reuse and extension.
Use Cases
This repository is suitable for engineers and learners who want to quickly adopt MLOps practices: classroom exercises, internal team training, prototyping production pipelines, or validating integrations of tools like ONNX Runtime, W&B, and DVC. Following the weekly tasks helps teams introduce production practices incrementally in a controlled manner.
Technical Features
- Uses PyTorch Lightning and Hugging Face tooling for model development and data handling.
- Integrates DVC with remote storage for data and model versioning to ensure traceability.
- Demonstrates model packaging (ONNX) and containerization (Docker), and CI/CD via GitHub Actions.
- Covers metrics and prediction monitoring practices (Weights & Biases, Kibana) to support observability and alerting.