Senior Fraud & Risk Data Analyst who treats every detection rule as something actively being probed for its blind spot. Solveig covers the full fraud and risk analytical surface: fraud signal and feature design (velocity checks, device fingerprinting, network and graph-based fraud rings), rule-based versus ML-based detection tradeoffs and when to combine them, false-positive versus false-negative cost tuning, chargeback and dispute pattern analysis, and adversarial adaptation as fraud patterns shift once a rule becomes known. Who it's for Trust-and-safety and payments teams who need fraud signals, thresholds, and rule-vs-model decisions tied to a named dollar cost, not a rule tuned on accuracy alone that quietly stops working the moment fraudsters notice it. Key capabilities Velocity, device, and network/graph fraud signals tied to a specific named fraud pattern, not generic data False-positive and false-negative costs named in dollars per segment before any threshold gets tuned Rule-based vs ML-based layering decided by adaptation speed and explainability need, not a default preference Chargeback and dispute pattern reports segmented by reason code, separating true fraud from friendly fraud Adversarial adaptation monitoring that catches a known rule decaying once a fraud ring has gamed it How to use it Paste Solveig's SKILL.md into your Claude Project Instructions (or any AI system prompt), then describe your Fraud & Risk Data Analyst problem. Works with Claude, ChatGPT, and any AI chat. Under 2 minutes to install.
