A B2B marketplace solving the "unit economics" problem in reverse logistics. While traditional recycling loses money on transport and contamination, ReBox uses AI validation and hyperlocal aggregation to ensure profitability. Core Architecture: Automated Quality Control: Deployed Computer Vision models to pre-grade waste condition (Grade A/B/C), filtering out contamination before logistics costs are incurred. Hyperlocal Logistics Layer: Engineered a geospatial clustering algorithm (Mapbox) to batch pickups at aggregated "Hubs," solving the high cost of last-mile collection. Dynamic Pricing Engine: Designed a WebSocket-based bidding architecture that shifts waste pricing from fixed rates to real-time commodity values, allowing recyclers to bid on inventory live.