Why MLOps Needs to Be Data-Centric

What is MLOps? Why MLOps needs to be data-centric

This is the beginning of a series of articles on the data-centric MLOps. The planned series is:

1. Why MLOps Needs to Be Data-Centric (current article)
2. Six Stages of Data-Centric MLOps
3. The Wheel of Data

Table of Contents

· The Origin
· What is DevOps?
· MLOps on the cheap?
· Software 2.0
· Model-Centric vs. Data-Centric Approaches
∘ The Model-Centric Approach
∘ The Data-Centric Approach
· Summary

Photo by Alexander Sinn on Unsplash

The word ‘MLOps’ became a hot keyword these days. The trend is poised to continue as AI takes more roles in the industry and society as a whole. The article is aimed at explaining what MLOps is and why the concern for data quality should be at the center of MLOps.

The Origin

The word MLOps is the combination of Machine Learning (ML) and Operations (MLOps = ML + Ops). It refers to the set of engineering practices to develop and operate machine learning models in production. In actuality, it involves three disciplines: Machine Learning, Data…