Inspiration
What it does
How we built it
Challenges we ran into
Accomplishments that we're proud of
What we learned
What's next for FieldSight
Inspiration
FieldSight was inspired by the need to make jobsite inspections faster and safer without turning vision findings into a passive dashboard. HVAC, electrical, and plumbing technicians need a clear next action when a fixture photo shows a safety issue.
What it does
FieldSight is an agentic inspection console for HVAC/electrical/plumbing jobsites. A technician selects a fixture photo; OpenCV 5 performs substantive image analysis using PPE vest, open panel, and missing-label heuristics. Those findings change what the agent does next: CLEAR, HOLD, or ESCALATE, followed by tools such as ticket.create and sms.draft. OpenCV output is part of the control loop, not just a visualization.
How we built it
The app combines Next.js and TypeScript for the console, Python/OpenCV for vision analysis, and AWS in the documented architecture. The Agentic Vision Award path is central: the vision result is passed into the agent policy that selects the operational outcome and follow-up tool.
What we learned and challenges
We learned that useful agentic vision requires explainable, deterministic signals and explicit policy boundaries. The main challenge was connecting imperfect image heuristics to safe workflow actions, so FieldSight makes the findings and resulting CLEAR/HOLD/ESCALATE decision visible. Technical report and architecture details are in docs/REPORT.md and docs/DEMO.md.
Built by Sam Desigan / Cubiczan.
Log in or sign up for Devpost to join the conversation.