marshmallow - Data Scientist (Credit) - Car Finance
Requirements
• 2–5 years' experience in decision science or data science within consumer lending or another regulated financial services environment. • Hands-on experience developing or materially refining credit scorecards or risk models (logistic regression and/or gradient boosting), including the practicalities: sample design, binning, reject inference, validation and monitoring. • Experience working with Open Banking or other alternative data in credit decisioning. • SQL user, and comfortable working with large, imperfect datasets. • A solid grounding in statistics and a good instinct for when a result is real and when it isn't. • Commercial judgement: you can translate a model output into a lending decision and explain the trade-offs to a non-technical audience. • Curiosity and ownership. You ask why, you dig until you find the answer, and you'd rather ship something and improve it than wait for perfection. • Motor finance or other secured lending experience. • Experience of working within a rapidly scaling lending environment. • Familiarity with UK consumer credit regulation and how it shapes model design and use. • Experience using LLMs to speed up analytical work. • You already know some of the tooling Marshmallow uses: Snowflake, Python, Claude, Linear, Notion
Responsibilities
• Build, validate and monitor application and behavioural scorecards, using bureau, Open Banking and internal data to improve how we assess new-to-UK and thin-file customers. • Design, test and refine our credit strategy: cut-offs, affordability rules, pricing tiers and limits, with a clear view of the trade-off between risk, volume and margin. • Run experiments and champion/challenge tests to prove that changes to policy and models actually work before we scale them. • Explore new data sources and alternative signals, and build the case for using (or not using) them in decisioning. • Support model governance: document what you build, evidence that it treats customers fairly, and help us meet our Consumer Duty, CONC and model risk obligations. • Work with our engineers to get models and rules into production, and use AI tooling to make yourself and the team faster. • Bonus scheme designed to reward high performance • Private medical insurance with Vitality, mental health support with Oliva • Personal learning budget and 2 dedicated L&D days a year • Monthly flexible benefits budget to spend as you choose • 25 days holiday plus bank holidays • 4 weeks Work From Anywhere per year • We are not able to offer visa sponsorship for this position. • Initial call with a member from our Talent Team (30 mins) • Past Experience interview with Hiring Manager (60 mins) • Onsite interview consisting of: • Technical interview with a couple of the team (60 mins) • Culture interview (60 mins) • Diversity of thought • We know the best ideas come from having different perspectives in the room - and we're committed to hiring fairly, regardless of background, identity or experience. If you see yourself in this role, we'd encourage you to apply.
Benefits
• prefer not to say • As part of our hiring process, we carry out standard DBS and Cifas background checks. These verify there are no ongoing criminal proceedings and support fraud prevention. If you have questions, speak to our Talent Acquisition team. • Use the tick box to confirm you've read and understood • We take privacy seriously at Marshmallow. Our Recruitment Privacy Notice explains how we process and handle your personal data. • Please tick the box to confirm you've read the notice and agree to us contacting you about opportunities for up to 3 years. • If you're responding to a data retention extension request, tick the box to confirm we can continue holding your data for an additional 12 months. You can ignore the note below. If you'd prefer to opt out, simply ignore the email. • Note: The consent period lasts for 3 years • Recruiting Privacy Policy
Apply in one click
Upload My Resume
Drop here or click to browse · Tap to choose · PDF, DOCX, DOC, RTF, TXT