Product Manager, Fraud Data

Plaid•New York, NY

23h•$190,800 - $262,800

About The Position

The Fraud team is building the largest network of trusted identities in the US. Our products span the full user lifecycle: Identity Verification and Monitor for onboarding and compliance, Layer for frictionless returning-user experiences, and Protect, our fraud intelligence platform, as the connective risk layer for all of it. Protect has more inbound demand than any product we have launched. The challenge is signal. We are building the data infrastructure that will make Protect the best fraud signal in financial services. This is a product role for someone who lives at the intersection of fraud expertise and data science. You will work directly with our data team to build and improve Protect's signal stack: the attributes, scores, and fraud vector intelligence that customers use to make real-time decisions. You need to understand fraud, know how to partner with data scientists as a peer, and be able to translate between what models need and what products should deliver.

Requirements

Real hands-on fraud prevention experience: fraud vectors, detection systems, risk signal evaluation. Not general fintech.
Direct experience collaborating with data scientists or ML engineers to ship data-driven products. You need to be able to talk to a DS about feature distributions, label quality, and model performance, not just read summaries.

Nice To Haves

Experience at a fraud vendor (SentiLink, Sardine, Socure, Persona, Incode) or a fraud operations team at a large fintech or bank
Background with ML model inputs/outputs: feature engineering, offline evaluation, and moving from experimentation to production
Specialization in a specific fraud vector (ATO, synthetic identity, or first-party fraud)

Responsibilities

Own the feature roadmap for Protect: what we build, how we validate it, and how we measure coverage and fill rates
Build and manage the labeling pipeline as a product, including customer retros, data partnerships, and the label feedback loop into model training
Define the product requirements for fraud vector scores, including signal selection, precision/recall targets, and customer-facing output design
Partner with data scientists and MLEs as a day-to-day collaborator, not just a requester. Translate between model needs and product requirements in both directions
Work with GTM and Protect PM on how attribute improvements and new scores translate into customer-facing value propositions.