Predictive Supply Chain Nervous System

Developed an enterprise-grade logistics prediction platform that assigns a real-time delay probability score to each active shipment using an XGBoost classifier trained on synthetic historical supply chain telemetry.

Feature engineering covers container weight, route distance, historical port congestion indexes, and weather pattern variables, giving the model multi-dimensional context for each risk assessment.

High-risk shipments are plotted on an interactive Folium geographic map, clustering anomalous routes for immediate visual identification by operations managers.

An integrated Gemini LLM triage agent automatically generates specific, actionable rerouting and buffering directives for flagged shipments, replacing the reactive crisis management process.

Includes a full synthetic telemetry generation pipeline in Python to produce training datasets that simulate complex real-world supply chain variables.