UUEIC.app
中文
v1.0 · Production

SiteEye

AI Construction Progress & Quality Inspection

Open Console →v1.0 · Production

Positioning

Homeowners can't see the job site, project managers burn out shuttling between sites, and quality issues surface only after the fact — the chronic disease of renovation and construction. With SiteEye, workers and supervisors simply upload phone photos; a VLM identifies the construction phase and trade, pre-screens quality issues (flagged for human review), and auto-generates illustrated daily reports homeowners actually understand. Vision inference runs self-hosted on our own RTX 5090 — job-site photos never leave the premises. AI does the triage; humans always make the quality call.

v1.0 · Production

Core Features

Photo-in, inspection-out

Workers and supervisors upload photos via WeChat Mini Program or H5 — no training needed, no change to how the site already operates.

Automatic phase recognition

The VLM identifies the construction phase and trade in each photo (plumbing & electrical / carpentry / painting…) and files it onto the project timeline automatically.

Quality issue triage

AI pre-screens common quality issues; every flag enters a human review queue. The product explicitly states: AI triage + human confirmation — AI never makes the final call.

Homeowner daily reports

Auto-generated illustrated timeline reports per project, so homeowners can see the site remotely — less friction, more trust.

Schedule-deviation alerts

Recognized actual progress is compared against the plan; deviations trigger alerts so project managers spend their time where it matters.

Multi-project dashboard

One dashboard across all active sites: phase distribution, report coverage, pending quality reviews — every job site on one screen.

Architecture

Capture
WeChat Mini Program / H5: job-site photo capture and upload
↓
API
FastAPI: upload intake, task orchestration, report generation, alert push
↓
Inference
VLM (qwen2.5vl) on our own RTX 5090, self-hosted — photos never leave the premises
↓
Data & Dashboard
PostgreSQL project data + Next.js multi-project dashboard and homeowner report pages

Tech Stack

  • WeChat Mini Program / H5 (capture)
  • FastAPI (server API)
  • VLM: qwen2.5vl on RTX 5090, private inference
  • PostgreSQL (project & inspection data)
  • Next.js project dashboard

Stage & Milestones

  1. 0M0 — Phase-recognition accuracy target met on 3 phases at a single site (audited against human labels)
  2. 1M1 — All active Ganghong SmartDecor projects onboarded
  3. 2M2 — First paying external construction team

AI Output Risk Notice

AI output risk level: quality judgments are Medium risk. Every AI triage result must be confirmed by a human reviewer; the UI explicitly labels results as “AI triage + human confirmation”.

Pricing Direction

  • Free during beta (invite only)
  • Bundled with SmartQuote, or standalone per-project subscription
  • Self-hosted licensing (enterprise construction edition) on the roadmap

Apply for Beta

SiteEye is in invite-only Early Access. Construction teams and renovation companies are welcome to apply.

Send Application Email