• We worked with the client to design a new AI-powered platform for demand forecasting and inventory optimization. The goal was to replace fragmented planning with a single, reliable system that could connect suppliers, warehouses, and logistics partners
• The solution began with building a unified data layer that integrated feeds from 20+ systems, including supplier portals, ERP, and WMS. To ensure data quality and accessibility, we prepared and consolidated these sources into a Lakehouse architecture based on Databricks, giving planners a single point of truth
• We developed a forecasting model using Catboost, retrained weekly to capture both seasonality and short-term demand shifts. These models generated SKU-level predictions that directly fed replenishment logic
• On top of forecasting, we built an optimization engine that recommended purchase orders and restocking plans. The engine was tightly integrated with ERP and WMS workflows so that logistics teams could act on insights without manual reconciliation
