Feature store and MLOps

Writen by Devil 9:37 PM - 2 Comments

MLOps is a DevOps extension in which the DevOps principles are applied to machine learning pipelines. Creating a machine learning pipeline differs from creating software, primarily due to the data aspect. The model's quality is determined by more than just the code's quality.

It is also determined by the quality of the data — i.e. the features — used to run the model. According to Airbnb, data scientists spend roughly 60% to 80% of their time creating, training, and testing data. Feature stores allow data scientists to reuse features rather than rebuilding them for each new model, saving valuable time and effort. Feature stores automate this process and can be triggered by Git-pushed code changes or the arrival of new data. This automated feature engineering is a crucial component of the MLOps concept.

ML Ops is the intersection of Machine Learning, DevOps and Data Engineering

• Facebook
• Tweeter
• Google+
• Whatsapp Pinterest