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

Core responsibilities

• Architect end-to-end ML systems for high-dimensional retail demand modeling
• Build production-grade Python-based ML pipelines
• Develop and test advanced forecasting approaches (hierarchical forecasting, gradient boosting, time-series models, hybrid ML approaches)
• Design allocation or constrained optimization logic to translate forecasts into model stock targets
• Validate model lift using backtesting and live pilots
• Work cross-functionally to translate ambiguous business requirements into measurable outputs
• Document methodology in a way that stands up to executive scrutiny