Gerotto

Gerotto wanted to validate, quickly and in a real operating context, an "AI Control Room" architecture: camera video streams feeding an AI agent that reads what's happening on site and raises events an operator can review.

Svapna built the prototype around motion detection on up to four channels — when movement is detected, three frames are captured and sent to Google Gemini to check whether people on site are wearing required safety equipment, producing a structured result and a reviewable event list on the Control Room page.

The prototype runs on a multi-tenant platform with token-based access, plus technical support for a Raspberry Pi device that reads RTSP/H.264 camera streams and forwards them to the cloud. It's a validation build rather than a production system — persistence, roles, auditing and governance are scoped for a follow-on phase that reuses these components.