We run a supply chain optimisation platform for mid-size manufacturers. Our forecasting models generate 72-hour demand predictions used directly in production scheduling decisions. Current architecture: Prophet as baseline, gradient boosting ensemble for short-horizon, no proper uncertainty quantification. We know the weak points. We need a senior data scientist who can take this stack seriously — evaluate it rigorously, introduce probabilistic forecasting where the business case justifies it, and own the model development roadmap through the next major product cycle.
Responsibilities
Own the forecasting model stack and its performance
Introduce probabilistic output layers to improve planning reliability
Design rigorous offline and online evaluation frameworks
Work with our domain experts to incorporate manufacturing constraints into model inputs
Mentor two mid-level data scientists on the team
Requirements
5+ years time-series forecasting in production environments
Deep Python expertise (Pandas, Statsmodels, LightGBM or XGBoost, and ideally Darts or GluonTS)
Experience with probabilistic forecasting and uncertainty quantification
Able to evaluate model performance with appropriate metrics (MASE, WAPE, coverage) and explain trade-offs
Familiarity with supply chain or manufacturing data is a strong plus
Benefits
Technical ownership of a complex, commercially critical forecasting system