Computer Vision Data Scientist, Receipt Fraud
Role details
Job location
Tech stack
Job description
-
Analyze large-scale receipt data for fraud patterns and anomalies.
-
Develop statistical methods to detect subtle inconsistencies in receipt data.
-
Design feature engineering strategies combining OCR, visual embeddings, and behavioral signals.
-
Build and optimize ML models for fraud detection using collected data points.
-
Develop fraud scoring algorithms that combine multiple detection signals and model outputs.
-
Implement threshold optimization strategies balancing precision and recall for different risk levels.
-
Design comprehensive fraud scoring systems.
-
Develop weighted scoring mechanisms adaptive to fraud types and retailer patterns.
-
Create interpretable scoring frameworks for manual review teams.
Requirements
Do you have a Bachelor's degree?, 4+ years as a data scientist with experience in fraud detection.
- Strong expertise in hypothesis testing, time series, and anomaly detection.
- Hands-on experience with classification, ensemble methods, and deep learning (scikit-learn, XGBoost, PyTorch/TensorFlow).
- Computer Vision - Strong experience with image processing and embedding, specifically EfficientNet and FAISS, is a plus.
- Experience with high-volume transaction processing and real-time decision systems.
- Knowledge of retail/e-commerce fraud patterns preferred.
- Familiarity with document fraud techniques and anti-fraud methodologies.
- Part-time commitment with flexible hours
