An excellent paper regarding the complex states of MLOps today: “Machine Learning Operations (MLOps): Overview, Definition, and Architecture” - by Dominik Kreuzberger, Niklas Kühl, and Sebastian Hirschl. What I like the most is the figures depicting the End-to-end MLOps architecture with different roles and their intersections contribute to the MLOps paradigm.
- MLOps (Machine Learning Operations) is a paradigm, including aspects like best practices, sets of concepts, as well as a development culture when it comes to the end-to-end conceptualization, implementation, monitoring, deployment, and scalability of machine learning products. Most of all, it is an engineering practice that leverages three contributing disciplines: machine learning, software engineering (especially DevOps), and data engineering.
- In the real world, we observe data scientists still managing ML workflows manually to a great extent.
- Organizational challenges: To successfully develop and run ML products, there needs to be a culture shift away from model-driven machine learning toward a product-oriented discipline.
- Operational challenges: These repetitive tasks yield a large number of artifacts that require a strong governance as well as versioning of data, model, and code to ensure robustness and reproducibility.