WPP
2025—presentData Scientist, Global Functions
I design machine learning systems for complex enterprise data: procurement classification, standardized analytics, supplier entity resolution and master data management.
WORK / APPLIED MACHINE LEARNING
From messy data
to useful systems.
My industry work centers on problems where modeling is one part of a larger engineering task: making enterprise data usable, forecasting financial activity and building systems that can operate beyond an experiment.
Sandoz
2023—2024Data Scientist, Global Functions
I developed and deployed deep learning models for financial forecasting, connecting time-series modeling with operational planning.
BIsiona
Earlier experienceData engineering & business intelligence
I worked on data engineering and business intelligence, including projects for pharmaceutical clients. This was the foundation for my move into applied machine learning.
WPP / ENTERPRISE ML
Finding structure
at the source.
Classifying procurement expenditure
I designed a production-scale deep learning classification system for procurement expenditure, alongside standardized classification approaches for enterprise analytics. The goal is a consistent representation of operational information that people can use across the organization.
Resolving supplier identity
I designed and implemented an enterprise master data management and supplier entity-resolution system. These problems sit at the intersection of modeling, data quality and the operational meaning of a record.
The engineering questions
How should a classifier be evaluated when categories overlap? When should two records be treated as the same supplier? What makes an output useful to the people who rely on it?
These are the kinds of questions behind the work. Employer data, internal architectures and undisclosed performance figures are kept private.
SANDOZ / FORECASTING
Models that meet
operational needs.
At Sandoz, I developed and deployed deep learning models for financial forecasting. It brought time-series modeling into a setting where predictions have to support practical planning.
My research in sequence modeling informs how I approach these problems, while deployment brings a different set of questions about reliability and use.
Technical practice
Python · PyTorch · Hugging Face · SQL · Databricks
Classification, entity resolution, time-series forecasting, model evaluation and production ML engineering.