https://youtu.be/9W2B6U5_htA

Project Overview

• Product: Offline-first visual field-inspection app; no AI currently runs inside the app
• Role: Product designer and AI-native builder orchestrating Codex and Claude Code across planning, implementation, review, and testing
• Development approach: Phase-by-phase AI-assisted development, automated tests, emulator validation, and physical-device testing
• Future direction: Optional post-inspection automation may use AI to transform completed reports and dictated notes into proposal-ready content, while the core field workflow remains deterministic and dependable

Introduction: Keeping Evidence Connected to Its Location

Groundmarker began with a problem I observed in crawlspace inspections: the inspector draws a property sketch, takes dozens of similar-looking photos, and later has to reconstruct where each photo came from.

Groundmarker preserves that context at the moment the evidence is collected. The inspector draws the inspected area, places a numbered marker where a problem is found, and takes photos directly through that marker. The app maintains the connection:

Sketch location → Numbered finding → Linked photos → Client-ready PDF

The first version is designed around crawlspace inspections, where poor lighting, gloves, limited movement, and unreliable connectivity make complicated data entry impractical. However, the underlying workflow can support other field inspections where visual evidence must remain connected to a physical location.

What I am testing: Can inspection documentation become a natural part of fieldwork instead of a separate organization task afterward?

1. The Field Problem

A typical inspection may involve:

• A paper sketch or separate drawing app
• Photos stored in the phone’s general camera roll
• Notes recorded somewhere else
• Manual photo sorting after leaving the site
• A report or proposal assembled separately

The inspector must remember how all these pieces relate.