Senior Data Scientist - Credit Risk Modelling

Hybrid in London or remote in the UK

We're looking for a Senior Data Scientist to join our Credit Risk Modelling team.

You'll shape how iwoca models credit risk – setting the technical direction on multi-quarter projects, and lifting the bar for the team as you go.

The company

Small businesses move fast. Opportunities often don’t wait, and cash flow pressures can appear overnight. To keep going, and growing, SMEs need finance that’s as flexible and responsive as they are.

That's why we built iwoca. Our smart technology, data science and five-star customer service ensures business owners can act with the speed, confidence and control they need, exactly when it's needed.

We’ve already cleared the way for 100,000 businesses with more than £4 billion in funding. Our passionate team is driven to help even more SMEs succeed, through access to better finance and other services that make running a business easier. Our ultimate mission is to support one million SMEs in their defining moments, creating lasting impact for the communities and economies they drive.

The team

The Credit Risk Modelling team owns credit risk and Customer Lifetime Value (CLtV) modelling for iwoca's UK and German lending. That covers the probabilistic machine learning models behind every credit decision, plus the CLtV models that shape pricing and portfolio strategy.

The team is around twelve data scientists. Some work on auto-space models that decide within five minutes from data candidates authorise us to pull. Others work on manual-space models that pick up when a credit analyst adds digital footprint and income data, which lands a decision within 24 hours.

The role

As a Senior Data Scientist in Credit Risk Modelling, you'll own credit risk and CLtV projects end to end – from spotting where the models hold the business back to getting the change live. The work ranges from keeping today's models accurate as conditions change, through incremental improvements, to research that could replace them.

You'll decide which problems to take on and how to solve them. With AI making prototyping cheap, more of the role now goes on research. You'll review other data scientists' modelling choices and coach newer members of the team.

Live examples of the work:

Causal estimation of offer terms. Estimating the causal effect of amount, duration, and price on customer outcomes.

Unifying automated and manual models. Finding a principled way to put the two model families, which optimise slightly different objectives, on a common cost function.

IFRS accounting model. Developing a multi-stage model where recovery predictions feed back to improve the upfront loss estimates that International Financial Reporting Standards (IFRS) require.

Generalising credit and CLtV. Exploring whether a single, more general model could replace the separate credit and CLtV models.

The requirements

Essential:

Applied statistics. You have a background in statistics from a quantitative field – typically a degree or Master's in Statistics, Mathematics, Physics, or similar. You've applied it to commercial problems for several years, and chosen the methods yourself. You know when the data breaks a method's assumptions, and what that does to your answer.

Probabilistic reasoning. You quantify the uncertainty in each estimate, and treat calibration as a first-order concern.

Production ML. You've built supervised ML models end to end – exploration, training, deployment, and monitoring.

Judgement. You critically evaluate model output – yours, a colleague's, or an LLM's – and can explain why a choice is right. You defend your reasoning under challenge, and challenge others' when the evidence points elsewhere.

Project leadership. You've led modelling projects you spotted yourself, from framing the question to a decision the business acted on. Some of that work changed how a team models a problem.

AI leverage. You've used AI to produce work you couldn't have delivered without it, and built tools or workflows your colleagues now use. You judge where it helps and where it doesn't.