Principal Data Scientist, Fraud Modelling at Impact

About the Role

• We're seeking a Senior Data Scientist specializing in Fraud and Risk to join our Cape Town Data Science team. In this role, you'll be at the forefront of protecting our affiliate marketing ecosystem by researching, developing, and deploying ML models that detect and prevent fraud across attribution, lead quality, and partner compliance. You'll work on high-impact problems spanning traditional fraud patterns and emerging threats—from attribution manipulation to browser extension abuse—while building production systems that scale. This is an opportunity to combine rigorous analytical work with tangible business impact in a fast-moving, adversarial domain.

Core Responsibilities

Research & model development

• Conduct R&D on fraud detection and risk monitoring across the digital advertising ecosystem, including attribution fraud, lead fraud, click injection, browser extension abuse (e.g., Honey-style coupon hijacking), brand safety violations, and creator authenticity verification.
• Design, prototype, and validate ML models and rule-based systems for fraud detection, partner risk scoring, compliance monitoring, and trust & safety workflows.
• Research and apply graph-based fraud detection techniques (community detection, link analysis, behavioral clustering) and explore graph database applications for modeling relationships between users, devices, transactions, and partners to uncover coordinated fraud rings and suspicious network patterns.
• Stay ahead of emerging fraud patterns through continuous learning—monitoring industry trends, reviewing academic literature, exploring data for novel anomalies, and collaborating closely with Product, Compliance, and Trust & Safety teams.

Production deployment & iteration

• Deploy Fraud and Risk ML models to production; own the end-to-end delivery from ETL, feature engineering, model training, deployment, to monitoring.
• Iterate on live models by adding new features, improving performance (precision/recall/F1), and reducing false positives.
• Partner with MLOps and Engineering to ensure models are robust, scalable, and production-ready (testing, alerts, drift monitoring, retraining pipelines).

Data analytics & insights

• Perform deep-dive analyses on fraud trends, partner behavior, and risk patterns to inform model strategy and business decisions.
• Translate analytical findings into actionable recommendations for Product, Marketing, and Finance stakeholders.
• Build dashboards and reports to communicate model performance, fraud impact, and risk metrics to leadership.

Cross-functional collaboration

• Work closely with Product, Engineering, Compliance, and Finance to scope requirements, prioritize work, and align on success metrics.
• Communicate technical work clearly to non-technical audiences; present findings and tradeoffs in planning forums and reviews.
• Contribute to a culture of experimentation, documentation, and knowledge sharing within the Data Science team.

Qualifications

Required

• Experience: 5+ years in data science, ML, or advanced analytics, with at least 2+ years focused on fraud detection, risk modeling, or anomaly detection in production environments.
• Fraud & risk domain expertise: Demonstrated experience building and deploying fraud or risk models (classification, anomaly detection, time-series analysis, graph-based methods).
• Technical skills:
  • Strong Python and SQL; proficiency with ML libraries (scikit-learn, XGBoost, LightGBM, or similar).
  • Experience with feature engineering, model evaluation (ROC/AUC, precision-recall, cost-sensitive learning), and handling imbalanced datasets.
  • Familiarity with production ML workflows (versioning, monitoring, A/B testing, model retraining).
• Analytical rigor: Strong foundation in statistics and ML; ability to design experiments, validate models, and interpret results with business context.
• Communication: Ability to translate complex technical work into clear insights for stakeholders; experience presenting to cross-functional teams.
• Education: Bachelor's in a quantitative field (CS, Statistics, Math, Engineering, or similar); Master's/PhD preferred.

Preferred / Nice to have

• Experience in affiliate marketing, ad tech, or e-commerce fraud (attribution fraud, click fraud, lead validation, coupon abuse).
• Familiarity with browser extension detection, fingerprinting, or device/user identity resolution.
• Experience with graph analytics or network-based fraud detection (community detection, link analysis, behavioral clustering).
• Knowledge of privacy-preserving ML techniques or working with privacy-constrained data.
• Experience with real-time or near-real-time scoring and low-latency deployment (e.g., REST APIs, streaming pipelines).
• Familiarity with GCP tools (BigQuery, Vertex AI, Cloud Run) and/or Databricks/Spark for large-scale data processing.
• Exposure to rule engines, decision trees, or hybrid rule-ML systems for compliance and risk workflows.