Asset Intelligence — Field Evidence Intelligence Platform | Built for Meta Global AI Hackathon
Every town has assets that can’t fail. Culverts, bridges, headwalls.
Manage Assets Inc is building Asset Intelligence — an AI field evidence intelligence platform for municipal maintenance coordinators.
How it works: Inspector uploads 1-5 photos + field note (EXIF GPS stripped). System returns SAM 3.1 masks + Muse Spark 1.3 vision fusion + Spark evidence-linked JSON + Wonder discrepancy/missing check. Human must accept/edit/reject.
Why it’s different: Industrial Forensic Ledger. Click a SAM mask on the photo, a 1px steel tether draws to the observation row, that row highlights, cited note span underlines. Every sentence traces to a pixel. No blur that hides a crack. Keyboard-operable chips, 32px+ glove targets, sunlight contrast #18181b on white, mono utility ui-monospace for mask_id, spans, cost.
Meta-only orchestration: SAM 3.1, Muse Spark 1.3 (primary + fallback), Spark reasoning, Muse Wonder 1.0 planning, Muse Image synthetic evals, Muse Code. FastAPI + Pydantic v2 fail-closed validator, JWT/API key, rate limiting, OTEL/Prometheus, cost metering. $0.07 estimated, never $0.00.
Try it: Live Demo on Hugging Face + GitHub public repo for the hackathon submission.
Built in Akwesasne, New York / Cornwall, Ontario by Monica Peters.
Disclaimer: Does not certify safety. Requires qualified inspection.
