Senior data scientist · Fintech & e-commerce · Amsterdam
Five years building and deploying production ML systems — from a fraud model that cut false positives by 38% to a churn pipeline serving 4.8M customers.
Case study
The fraud scoring model he rebuilt at Adyen, from business question to outcome.
Reduce card fraud without increasing the false positive rate.
Rebuilt the fraud scoring model with gradient boosting, using 140 engineered features on 18 months of transaction history.
False positives down 38% and fraud losses down 22% — a saving of €2.1M a year.
Models & analyses
Infrastructure, models and reporting across Adyen, Jumia and Equity Bank.
A feature store in Python and Redis serving model inference in real time.
p99 under 8ms at 14,000 TPS
An experimentation framework adopted by three other data science teams.
40% faster experiment cycles
Identifying customers likely to churn within 90 days, across a 4.8M customer base.
Retention campaign ROI of 3.4x in the first quarter
A demand forecasting model covering 12 product categories.
Overstock down 18%, stockouts down 24%
Dashboards used daily by 14 credit officers.
6 hours a week of manual reporting removed
A SQL-based pipeline for regulatory reporting.
Monthly close cut from 3 days to 6 hours
Projects & publications
Technical stack
A data scientist who turns ambiguous business problems into reproducible, production-ready analytical solutions.
Tariq is a Senior Data Scientist at Adyen in Amsterdam, where he works on fraud scoring, real-time feature serving and experimentation. Before that he built churn and demand models for Jumia Technologies (Nairobi, remote), and started as a data analyst at Equity Bank.
He has mentored two junior data scientists, both promoted to mid-level within 14 months, and at Jumia established the first data science notebook standards and reproducibility guidelines for a team of eight analysts.
Experience
Adyen N.V., Amsterdam
Fraud scoring model (€2.1M a year), real-time feature store, A/B testing framework; mentored 2 junior data scientists.
Jumia Technologies, Nairobi (Remote)
Churn prediction on a 4.8M customer base, demand forecasting for 12 categories, notebook and reproducibility standards.
Equity Bank, Nairobi
Credit risk dashboards in Tableau and a SQL pipeline for regulatory reporting.
Education
University of Cape Town
Distinction. Thesis: predictive modelling for mobile money fraud in sub-Saharan Africa.
University of Ghana
First Class Honours.
Contact
tariq.osei@email.com
Fictional candidate — contact details shown as plain text only.