Geospatial Data Scientist: Fuel-Retail Demand & Site-Selection Model

I need a spatial data scientist to build the predictive model behind a fuel-retail network planning platform for the UAE. The UI, data pipelines, and product layer already exist. The model does not. You build it end to end. What you build: Demand and revenue model: predicted fuel volume and revenue in real units for any location Cannibalization model: demand impact of a new site on an existing fleet of ~600 stations Site scoring and white-space detection Calibration and back-testing against 2 years of station transaction data (target 85-90% accuracy) Model served via API (e.g. FastAPI) for the front-end to consume Data you will work with: GPS mobility, demographics on a 250x250m grid, competitor POIs, real-estate costs, and EV/growth forecasts. Ground truth: station list and historical fuel and C-store transactions. Required: Proven spatial demand or site-selection modelling (fuel retail, retail/QSR, telco coverage, or quant real estate) Gravity / Huff / competing-destinations models for trade-area and market-share allocation Predictive ML: regression and gradient boosting (XGBoost, LightGBM) You understand why random cross-validation fails on spatial data and how to back-test correctly Geospatial Python: GeoPandas, shapely, PySAL or rasterio Notes: Front-end and dashboards are out of scope, already covered Work runs inside a secured Azure UAE North environment (data stays in-region). Remote is fine ~4 month contract, with a likely 30-month follow-on if the pilot is approved When you apply, do not send a generic ML pitch. Tell me about one spatial demand or site-selection model you have actually built: what it predicted, how you validated it, and the accuracy you achieved. ---- Extra details in the attached PDF