Real-time visibility for a nationwide logistics fleet

The challenge

Dispatchers coordinated hundreds of daily shipments through spreadsheets and phone calls — no live view of fleet status, and no way to warn customers about delays before they happened.

What we built

We designed and shipped a real-time operations platform: live GPS ingestion over Kafka, a dispatcher command center built in Next.js, and ML-based ETA predictions that flag at-risk deliveries hours in advance.

How we did it

1. 01

   One-week scoping sprint with dispatchers on the floor — the roadmap was built around their actual failure points, not a feature wishlist.

2. 02

   Stood up streaming ingestion over Kafka for live GPS pings from hundreds of vehicles, with schema contracts so bad data fails loudly.

3. 03

   Shipped the dispatcher command center in week eight: live fleet map, exception queue, and one-click customer notifications, built in Next.js.

4. 04

   Trained an ETA prediction model on two years of delivery history, weather, and traffic — surfacing at-risk deliveries hours before they slipped.

5. 05

   Ran old and new workflows in parallel for a month; dispatchers switched voluntarily because the platform was simply faster.