Express Logistics Player Eliminated POD Data Entry — 4.2M Docs/Year Automated

The Challenge

POD (Proof of Delivery) data entry was a 6-person team manually typing fields from photographed PODs uploaded by drivers. Backlog was always 24-48 hours, blocking invoicing and ageing customer disputes. Photo quality varied wildly — dust, low light, partial smudges.

Our Solution

Iedeo deployed an OCR + LLM pipeline triggered when drivers upload POD photos through the driver app. Fields auto-extract, route to TMS, generate invoices. Low-confidence cases (<88%) drop into a 1-person review queue. ANPR added later for gate vehicle ID.

Architecture & Stack

• Driver app (React Native) with offline POD capture
• PaddleOCR for typed fields + Azure Form Recognizer for handwriting
• GPT-4o-mini for field extraction + fuzzy customer matching
• YOLOv8 ANPR for Indian plate variations at gates
• TMS integration (custom + LogiNext)
• PostgreSQL audit log, S3 image archive (90-day retention)

Timeline

2 weeks discovery → 7 weeks build → 4 weeks pilot on Mumbai-Bengaluru lane → 3 weeks national rollout

“Our finance team used to chase PODs. Now they chase nothing. The invoicing cycle dropped from days to hours.”