Ai-Robotics

1. Background & Problem Statement Manual or rules-based inventory planning is reactive and error-prone. Common impacts: 10–20% excess inventory 5–10% lost sales from stockouts Existing ERP systems lack predictive intelligence. 2. Objectives Reduce excess inventory by ≥15% Reduce stockouts by ≥20% Improve inventory turnover and cash flow Enable data-driven replenishment decisions 3. Scope In scope Demand forecasting (daily/weekly) Replenishment recommendations ERP integration (read/write) Dashboard and KPI reporting Out of scope Supplier renegotiation Pricing optimization Hardware changes 4. Technical Approach Architecture Data ingestion from ERP → Feature store → ML forecasting service → Replenishment engine → ERP Models Baseline: SARIMA Production: Gradient Boosting / LSTM (SKU-level) APIs REST (JSON) Batch + near-real-time sync Complexity Forecasting: O(n × t) where n = SKUs, t = time windows Trade-offs Accuracy vs. interpretability Model complexity vs. operational stability