AI/ML Fraud Solution Engineer

About the role

This role will be responsible in building an AI/ML-powered fraud prevention program that combines proven machine learning for detection (gradient boosting, graph neural networks, sequence models) with emerging agentic AI and generative AI for analyst assist, case triage, and adversarial testing. This role designs and builds those systems end-to-end.

This is a hands-on engineering role. You architect the solution and you write the code. You need to understand fraud enough to have opinions on model design; you need to understand modern AI stacks enough to have opinions on architecture.

Key responsibilities

• Build real-time scoring models for transaction fraud detection
• Design and implement application fraud detection for onboarding and lending applications
• Develop graph neural networks over customer-to-customer transfer patterns to identify mule rings before they operationalise
• Build LLM-powered agents that autonomously gather context around fraud alerts and produce structured case summaries for fraud analysts
• Conduct adversarial red-teaming and synthetic fraud pattern generation for testing detection systems
• Architect and implement end-to-end AI/ML systems across the full lifecycle: feature engineering, training, evaluation, serving, monitoring, and retraining

About you

• Bachelor's or Master's degree in Data Science, Computer Science, Statistics, Computer Engineering, or a related field
• 5+ years in ML/AI engineering, with production deployment experience
• Deep experience with either fraud detection models (gradient boosting, anomaly detection, sequence or graph models) or modern AI stacks (LLMs, agentic frameworks, GenAI applications)
• Python fluency; strong with at least one deep learning framework (PyTorch, TensorFlow)
• Comfortable with real-time inference systems, model serving, and MLOps
• Understanding of the adversarial nature of fraud: models degrade over time, adversaries adapt, monitoring is core to the job
• Comfortable working in a regulated environment where explainability is a design constraint
• Experience with graph neural networks (for mule detection) or sequence models (for transaction fraud) strongly preferred
• Production experience with LLM-based agentic frameworks (LangChain, LangGraph, or equivalent custom agent systems) strongly preferred
• Prior experience in a bank, insurer, or payment processor strongly preferred