For models that need to operate in real-time, the biggest roadblock stopping models from working in production are the complex data pipelines needed to deliver data in real-time. Fraud detection is frequently one of the most demanding real-time models, requiring real-time decisions made on real-time data inputs.
Feature stores are built to help solve the challenge of delivering data to production models. With a tool like Tecton, as soon as you’re done training your model, your data is available to be served in real time.
In this workshop, we will demonstrate how you can use a feature store to build new features for fraud detection, build a training dataset, use that dataset to train a model, then deploy that model with AWS SageMaker.
