• PayPal uses declarative, config‑based feature engineering to streamline ML workflows. • Data scientists declare feature specs, leaving construction to engineering teams. • Approach reduces Time to Market, enabling faster production releases. • Cost analysis shows reuse of existing features cuts TCO dramatically. • Features categorized into three complexity layers for tailored strategies. • Declarative model supports real‑time fraud detection across 400M users. • First of a two‑part series detailing scale, TTM, and TCO benefits.

Article Summaries:

  • PayPal has adopted a declarative, config‑based feature engineering approach to streamline fraud‑prevention machine‑learning workflows. By letting data scientists declare feature specifications instead of coding construction logic, the company separates feature design from execution, enabling engineers to manage scaling, time‑to‑market (TTM), and total cost of ownership (TCO) more efficiently. Features are categorized into simple, code‑based, and analytical tiers, with reusable simple features prioritized to reduce duplication and cost. The initiative, first detailed in a two‑part series, aims to improve predictability of TTM, lower maintenance overhead, and support PayPal’s real‑time fraud detection at scale.

Sources: