Data Science & BI ConsultantSep 2024 – now Cérélia (Meravit client) · Paris2 yrs 2 mos A Meravit client engagement: proposed and shipped AI demand forecasting used weekly across 14 countries.
- Designed a hybrid ensemble forecasting model in Python (linear-regression trend + horizon-specific XGBoost models on residuals) delivering a 72-week rolling forecast for 815+ SKUs in 14 countries
- Raised 6-month weighted forecast accuracy from ~73% (existing SAP forecast) to 82% (+9 pts)
- Deployed the model to production on Microsoft Fabric with weekly runs and automated output into SAP, coordinating up to 7 internal and external stakeholders
- Built 10+ Power BI reports on a star-schema model (DAX, Power Query M), used by 300+ people in 6 departments: Sales, Finance, Controlling, Operations, Customer Service, Engineering
- Aligned commercial and operations teams on a single forecast, giving Operations control over AI use per SKU–client