Lead Data Scientist - AML Handling

About this role

Lead Data Scientist - Anti-Money Laundering (AML) Handling & Prevent We’re looking for a Lead Data Scientist (IC3) to join our Anti-Money Laundering (AML) Handling & Prevent team in London. This role is a unique opportunity to work on the intelligence system at the core of our operational handling and prevention work. You'll help automate components of our operational systems, establish robust LLM evaluation pipelines, and build solutions that reduce financial crime risk.

What you build will have a direct impact on Wise’s mission and millions of our customers. The AML Handling & Prevent team offers an exciting environment for applying cutting-edge Generative AI solutions and machine learning architectures. This team is dedicated to enhancing our financial crime mitigation operations through advanced tooling, automated evaluation frameworks, and prompt optimization, aiming to streamline reviews and simplify the work of our operations staff.

As a Lead Data Scientist (IC3), you will drive technical strategy across handling and prevent initiatives, architect robust AI systems, establish post-deployment monitoring, and lead complex automation initiatives to reduce financial crime risk across Wise. Here’s how you’ll be contributing: End-to-End Automation & EDD LLMs: Lead the development and deployment of AI models designed to augment operational workflows (e.g.

Business and Consumer EDD LLMs), specifically targeting the automation of case summaries, red flag generation, risk classifications, and document requests. Evaluation Framework & Labeling Taxonomy: Establish labeling taxonomies and guidelines with EDD SMEs, construct evaluation datasets, and implement automated eval harnesses to systematically measure accuracy, precision, recall, and failure modes. Prompt Optimization & Experimentation: Audit existing prompts and run structured experiments (few-shot, chain-of-thought, context ordering) within a hypothesis-driven testing framework to reduce hallucinations, formatting errors, and prompt drift.

Shadow Testing & Monitoring: Design shadow mode deployments and parallel execution testing to safely evaluate prompts at scale, while implementing post-deployment monitoring for data and output drift. Full-Stack Deployment: Take ownership of the production pipeline by writing and deploying production-ready Python services. You must be willing to bypass engineering bottlenecks to ship value quickly while maintaining code quality.