Data Scientist (Fraud Prevention & Creditworthiness Decision Engine) - PT/FT - EMEA

Engagement: 4–8 weeks, with potential extension. 25–30 hrs/week preferred (40 hrs/week considered if candidate has strong full-time-equivalent output history)
Location: Central European Time ±a few hours (Europe, Middle East, Africa).
Reports to: CEO directly.

Overview

The company currently relies on a third-party payment provider's fraud protection. Moving to in-house BNPL removes that safety net, so this role owns building the full decision engine from scratch — not just architecting it.

Responsibilities

• Design and build an end-to-end fraud + creditworthiness decision engine for a new in-house BNPL product
• Build the risk/credit scoring model itself — candidate can choose the approach (statistical, gradient-boosted trees, etc.) based on what fits the data and use case
• Handle edge cases third-party tools don't cover well (e.g., no credit history but high fraud risk, or high fraud score with high creditworthiness)
• Pull, clean, and structure the underlying transaction/customer data needed to train and run the model
• Integrate third-party data sources (credit bureau, KYC/identity checks, etc.) where useful or mandatory — candidate can recommend which providers/APIs fit best for the German/EU market
• Build the model so decisions can be explained on request — relevant for EU adverse-action/consumer-credit requirements; exact method left to candidate's judgment
• Advise the company on what additional data/signals competitors use, with the goal of building something better than current market standard — not just parity
• Work directly with the in-house engineering team for implementation (they handle integration into the web app)

Must-haves

• Proven experience building fraud prevention and/or credit risk systems from zero — not just operating on top of existing infrastructure. Prior work at BNPL companies (e.g., Klarna) or startups building this from scratch is a strong signal
• Strong Python and SQL
• Experience building or scoring models for real-time/checkout-speed decisions (not just batch/offline scoring)
• Comfortable building/exposing APIs for engineering handoff
• Track record of autonomous delivery — no one is scoping this day-to-day

Nice-to-have

• Experience with German/EU credit bureau data (e.g., Schufa) or EU consumer lending compliance
• Familiarity with BNPL-specific risk models and competitor benchmarks

Note on tooling: specific modeling techniques, libraries, and third-party providers are left to the candidate's discretion — the company wants their recommendation on what's fit for purpose, not a prescribed stack.