BlogsStripeML Feature Development Platform

ML Feature Development Platform

ML Feature Development Platform

2
posts
2023–2024

This feature thread tracks Stripe's development of systems and platforms to scale Machine Learning (ML) feature development. The initial post introduces Shepherd, a system that adapted Chronon to address the challenges of scaling ML feature development at Stripe. It details how Shepherd enables efficient creation, management, and deployment of ML features, focusing on technical aspects like data pipelines, feature stores, and model training infrastructure. This capability aims to accelerate the ML development lifecycle. This post details the technical decisions and lessons learned in building Stripe Radar, focusing on its machine learning components and fraud detection capabilities, highlighting the evolution of their fraud prevention solution.

2024

Shepherd: How Stripe adapted Chronon to scale ML feature development

4/15/2024

This post introduces Shepherd, a system built at Stripe that adapted the Chronon framework to scale ML feature development. It details the technical challenges of managing ML features at Stripe's scale, including data versioning, feature lineage, and efficient feature computation. Shepherd's architecture and implementation are discussed, highlighting how it integrates with existing ML infrastructure to streamline the feature development lifecycle from experimentation to production.

2023

How we built it: Stripe Radar

3/29/2023

This post details the technical decisions and lessons learned in building Stripe Radar, focusing on its machine learning components and fraud detection capabilities, highlighting the evolution of their fraud prevention solution.