Staff Applied ML Engineer - Financial Crime

Wise is hiring a full-time Staff Applied ML Engineer for its Risk ML team in London, which sits within the FinCrime organisation. The starting salary is listed as £145,000 to £182,000 plus RSUs. The team is developing the newest generation of financial-crime detection, drawing on methods such as deep learning, GNNs and foundation models to catch more advanced fraud and laundering patterns.

The role is greenfield, and the engineer will set the architectural direction for Wise's use of modern machine learning in risk work, deciding which model families, serving patterns and training approaches to use. Responsibilities include designing and shipping sequence, graph and attention-based models that make decisions in real time across Wise's volume, and creating a reusable end-to-end path, running from trials through training to production, that future models can follow. The engineer tests foundation-model and embedding methods for representing transactions, works with data scientists on model evaluation, experiment design and causal measurement where clean A/B testing is not possible, and mentors engineers and data scientists in ML fundamentals and production practice. The Risk ML team has three areas, Risk Modelling, Feature Platform and Learning Loop, and this engineer sits in Risk Modelling, working with a lot of independence and few layers of management.

The posting asks for production work putting deep learning models into service that handles live traffic within tight latency limits, and the ability to make architecture decisions independently and explain the trade-offs. Candidates should understand optimisation options such as quantisation, precomputed embeddings and batching, have a strong grounding in deep learning, and show a record of shaping technical direction across teams. Proficiency in Python and PyTorch or similar tools, distributed training and ML pipeline orchestration are needed. Experience in fraud, AML or regulated financial services, graph-based methods in production, fine-tuning foundation models or evaluating LLMs, and building modern ML practice in scaling organisations are nice to have.