Hayden WittHayden Wittget in touch
← all projects
Computer Vision · Savi Solutions

Flow Board

A real-time drive-thru monitor on a $300 edge device — where leadership thought real-time on-site wasn't feasible.

stackRaspberry PiAI acceleratorEdge modelCMS signage
problem

The Customer Flow data was great for above-store reporting, but employees and district managers needed real-time visibility. Internally, leadership — head of engineering, CTO, CEO — believed real-time on-site wasn't feasible.

approach

Based on my own research, I believed a local model on a Raspberry Pi with an AI accelerator could work. I vibe-coded a rough prototype that validated we could do simple detections, use them to trigger timers, and later add a tracker into the model. I then designed the user experience and solved the on-site display problem using a signage stick that loads our dashboards via CMS — cost-effective and performant.

impact

A real-time solution at a fraction of the cost. Competitors run ~$10,000 per install plus hundreds per month; our two-device setup is ~$300–400 per install with much lower monthly costs. Locations see 10–30 seconds shaved off average service time, and the real-time data lets top brands like Swig run performance-based manager competitions.

stack

Local detection model on Raspberry Pi + AI accelerator, with a CMS-driven signage stick for in-store display.

demo
New Flow Board
Current Flow Board
QR code sign-in