Summary

• • What you will own You will design and build a repeatable, production-grade system that determines optimal model stock distribution at the store-SKU level for a national brick-and-mortar retailer (1,200+ stores, ~300 SKUs in category). The current allocation process begins with a store-tier level forecast. It ignores critical multi-dimensional signals that drive localized demand variation. Your mandate is to break that limitation. You will architect and implement a machine learning-driven optimization framework that ensures the right product mix lands in the right store — maximizing sales while reducing stockouts and misallocation. This is not theoretical research. You will: Translate business objectives (stockouts, % stock vs model stock, sales lift) into measurable optimization targets Design multi-dimensional forecasting models Engineer feature pipelines that incorporate regional, demographic, climate, historical sell-through, promotional, and behavioral signals Build and validate allocation or optimization logic on top of demand forecasts Deploy containerized solutions into production Run controlled experiments to prove incremental lift Iterate fast when reality disproves assumptions You will operate as a senior individual contributor with architectural authority. What success looks like Initial Stock (Forecast Based) model that materially outperform current baseline models A repeatable pipeline that retrains, recalibrates, and adapts Quantifiable reduction in stockouts Increased sell-through due to better SKU-store alignment A system that merchandising and supply chain teams trust