What is a feature store?
A feature store is the system that computes and serves the features a model uses, in training and in production, from one definition. It exists so the values a model learns from match the values it reads when it runs.
The idea
A machine learning model reads features: the inputs it was trained on, like how often a customer bought in the last month or how recently they logged in. A feature store is the system that computes those features, stores them, and serves them to the model when it runs. It sits between raw data and the model.
What it holds
A feature store keeps two copies of the same feature. The offline copy is the long history used to train a model. The online copy is the current value served to the model in production. Both come from one definition, so the feature means the same thing in both places.
Why teams use one
Without a feature store, each model computes its own features, and the same idea gets defined twice: once in a training notebook, once in production code. They drift apart, and the model degrades without anyone noticing. A feature store gives the feature one definition and one owner.
See it on your own traffic
Add the SDK, define one attribute, and see an attribute updating against your own traffic in the first session. The free tier runs on the real engine: no card, no sales call.