AI Transaction Graph Anomaly Detector

Gradrix Works

Remote · $130k – $160k / year · Full-time · Posted 4 months ago

This role is no longer accepting applications

Don't miss the next role like this

Create a free profile with your skills. We match you to new AI roles as they open, and companies hiring can contact you directly.

Create free profile, 2 minutes

Similar open roles you can apply to now:

See all open AI jobs

This role is no longer accepting applications

Don't miss the next role like this

Create a free profile with your skills. We match you to new AI roles as they open, and companies hiring can contact you directly.

Create free profile, 2 minutes

Similar open roles you can apply to now:

See all open AI jobs

About the role

Transaction fraud that bypasses per-transaction models is usually graph fraud: a ring of accounts connected through shared devices, addresses, or referral chains that individually look clean. The signal lives in the relationships, not the individual records. Building the graph layer that exposes those relationships — and running anomaly detection against it in near real time — is the capability gap our most sophisticated financial clients keep asking us to fill. Our fraud intelligence platform processes $11B in monthly transaction volume across 30 banks and payment processors. The existing models operate per-transaction. We've built the transaction graph — 8 million nodes, 60 million edges — but we haven't yet built the anomaly detection that turns graph structure into actionable fraud signals. You'll design the graph anomaly detection system: engineering structural features from the transaction network, identifying subgraph patterns that correlate with confirmed fraud rings, and building the real-time scoring pipeline that flags high-risk subgraphs for the fraud operations team. The technical challenge is latency: graph queries that take 3 seconds in batch can't take 3 seconds in a real-time decision flow.

Fully remote.

Responsibilities

Design graph structural features capturing entity relationship patterns correlated with fraud ring behavior
Build subgraph anomaly detection identifying suspicious network motifs and unusual connectivity patterns
Implement near-real-time graph scoring meeting latency requirements for transaction decision flows
Develop ring detection models learning from confirmed fraud ring structures in historical data
Create fraud signal integration feeding graph anomaly scores into the existing decision engine
Build evaluation framework measuring ring detection precision against confirmed fraud investigation outcomes

Requirements

6+ years in ML engineering with graph analytics or financial fraud detection experience
Strong Python, SQL, and Spark skills with graph data processing at scale
Experience with graph neural networks or graph-based anomaly detection
Deep understanding of payment network structures and fraud ring coordination patterns
Track record delivering fraud detection improvements with measurable loss reduction