Data scientist

Junior Data Scientist — Fraud & Risk

We are hiring a junior data scientist to help us detect and stop payment fraud in real time. You will work alongside our Data Lead and a small data team to build, deploy, and monitor production fraud risk models that protect both our customers and our payment volume.

Why this role exists

Fraud losses are a direct hit to our economics, and the current detection approach is fragmented across systems and rules. We need a dedicated data scientist focused on this problem — someone who can ship a baseline fraud scoring model, learn from production behavior, and iterate quickly. Your work will compound: every percentage point of fraud loss reduction (measured in bps of payment volume) goes straight to the bottom line, while keeping false positives low protects the legitimate customer experience.

What you'll do

Fraud model development. Build supervised classifiers on confirmed-fraud labels using gradient-boosted trees and, where it earns its place, deep learning. Focus on imbalanced classification techniques, careful feature engineering on transaction data, and honest evaluation (precision/recall, $ losses prevented).

Production deployment. Take models from notebook to staging to production. Partner with engineering on the scoring path, latency budgets, and the data contracts your features depend on.

Monitoring and iteration. Instrument deployed models for drift, score distribution shifts, and label feedback loops. Maintain dashboards that show the risk team how the model is performing on real money.

Cross-functional partnership. Work with risk, product, and compliance to make sure model decisions are explainable, auditable, and aligned with regulatory constraints.

What we're looking for

Strong Python — you can write clean, testable code, not just notebooks.

Solid SQL for pulling and shaping transaction data from a fragmented warehouse.

Applied machine learning fundamentals: gradient-boosted trees, imbalanced classification, proper train/validation splits on time-series data.

Exposure to deep learning and a clear sense of when it's the right tool versus when a simpler model wins.

Curiosity about fraud and payments — you want to understand how money actually moves, not just the model metrics.

Comfort working in a fragmented data environment without waiting for someone else to clean it up.

Nice to have

Experience with big data tooling (Spark, BigQuery, or similar).

Strong exploratory data analysis (EDA) habits — you investigate before you model.

Prior exposure to fintech, payments, or risk modeling.

Familiarity with feature stores and model monitoring.

What success looks like

Within your first deliverable window, you ship a baseline fraud scoring model deployed to staging, with a clear evaluation report on precision/recall against confirmed-fraud labels. From there, success means a measurable reduction in fraud loss rate (bps of payment volume) and a controlled false-positive/decline rate, while holding or improving auth approval rate.

Logistics

Fully remote within Egypt. Full-time, EGP 15,000–25,000 per month. Reports to the Data Lead within a small (2–5) data team in Engineering. Application deadline: 30 June 2026.

Benefits

Health insurance

Social insurance

Annual learning budget for courses, books, and conferences