Revolut
Lead Data Scientist • Revolut Business
London • Mar 2023 – present
Revolut Business offers borderless business accounts for companies and freelancers.
- Led a team of 3 data scientists that delivered ML projects for card fraud (5x block rate decrease; £20k/wk saved in chargebacks), receipt matching (migration to an in-house CV model), and LLM-based customer feedback categorisation
- Led BI migration to a data lake backend that reduced dashboard loading times by 95%
- Built a framework for A/B testing and tracking the performance of marketing comms
- Implemented and led the quarterly KPI review process for 20+ teams in the department
- Upskilled 30+ data analysts by teaching them linear models in an internal ML course
Senior Data Scientist • Revolut Business
London • Jul 2021 – Feb 2023
- Managed the central B2B data team, grew it 2x to 10 DA / DS / DE, and implemented product team embeddings, facilitating the support of 15+ product teams with analytics
- Prototyped and passed CTO's system design review for a company-wide A/B testing framework, later adopted by 20+ teams in the company (Python / Airflow / Trino)
- Led the migration of 60+ dashboards from Metabase to Looker (SQL / LookML)
- Mentored 5 data scientists / analysts across Revolut as their functional manager
Data Scientist • Revolut Business
London • Jul 2020 – Jun 2021
- Designed an ML-based system for automated receipt parsing & matching in expense management, resulting in a 2x increase of the product retention rate (Python / GCP)
- Built ETL pipelines driving the migration of Brexit-impacted businesses to a new legal entity, saving £10m+ in annual revenue that would be lost otherwise (SQL / Airflow)
- Built a family of ML models for customer LTV; productionised them in Python & Airflow (digital marketing optimisation) / in Java for online predictions (KYC ticket prioritisation)
- Built a system for collection, anonymization, and statistical analysis of customer NPS & feedback, driving the product roadmap and 6 KPIs (Python / spaCy / Airflow / Looker)
- Built a production-grade library for robust feature computation (Python / SQLAlchemy)
- Built time series forecasting models for business signups and card spend (FB Prophet)