Staff Applied ML Engineer - Financial Crime

Description

The company moves billions of dollars' worth of money across borders each year, and each transaction needs a decision on whether it is safe. The company's ML systems make that call at scale, in real time, in every market it operates in.

The Risk ML team is building the next generation of financial crime detection. It is investing in deep learning, graph neural networks and foundation models to catch more sophisticated fraud and money laundering patterns.

This is a greenfield role. You will set the direction for how the company applies modern ML to financial crime risk, with strong backing and involvement from senior leadership.

The company is a global technology company that helps people and businesses send, spend, and manage money internationally.

Responsibilities

• Design and ship deep learning and ML models for financial crime detection, including sequence-based, graph-based and attention-based approaches, serving real-time decisions.
• Define the architecture strategy for applying modern ML to risk, covering model families, serving patterns and training approaches.
• Build a reusable end-to-end pipeline, from experimentation through training to production, that future models will follow.
• Evaluate and prototype foundation model and embedding approaches for transaction representation across FinCrime domains.
• Work with Data Science on model evaluation, experiment design and causal measurement, including in domains where standard A/B testing is not always possible.
• Mentor engineers and data scientists on ML fundamentals, production practice and architecture decisions.

Requirements

• Experience in FinCrime, fraud detection, AML, or regulated financial services.
• Experience running graph-based methods such as GNNs, entity resolution or link analysis in production.
• Experience fine-tuning foundation models or evaluating LLMs.
• Experience introducing modern ML practices in organisations that are scaling their ML capabilities.
• Production experience shipping deep learning models that serve real traffic under latency constraints.
• Ability to make architecture decisions on your own, covering model selection, training infrastructure and serving, and to explain the reasoning and tradeoffs.
• Experience designing ML systems with hard latency and throughput targets, using techniques such as quantization, precomputed embeddings and batching.
• Strong grounding in deep learning, including gradient dynamics, attention mechanisms, graph message-passing and sequence modelling.
• A track record of shaping technical strategy across teams.
• Python, PyTorch or an equivalent framework, distributed training, and ML pipeline orchestration.

Conditions

Risk ML sits within the FinCrime organisation and owns the ML and AI foundation for financial crime detection. It is scaling into three pillars: Feature Platform, Learning Loop and Risk Modelling. You will be in Risk Modelling, working with data scientists, platform engineers, product managers and domain experts.

The team works with high autonomy and low hierarchy. You will own problems end to end, from research and architecture through production deployment and impact measurement. The team values engineers who shape direction, not only those who complete tickets.

The role is hybrid, based in London.

The company welcomes applicants from all backgrounds and is committed to equal opportunity.

Benefits

• Starting salary of £145,000 to £182,000, plus RSUs.
• Access to the company's benefits package.