Lawn Estimator

The lawn care industry is a $130B market where the sales process hasn't changed in decades. A homeowner wants a quote. They schedule an in-person visit. A landscaper drives to the property, walks around with a measuring wheel, drives home, and emails a number. Half those visits don't convert. The homeowner waited three days for a number they could have gotten in seconds — because the data to calculate it is literally photographed from space every few weeks.

Enter an address. The system pulls satellite imagery, runs SegFormer ML semantic segmentation to identify grass vs. hardscape vs. structures, calculates treatable area in square feet, and produces an instant cost estimate. The pipeline: address → geocoding → satellite tile fetch → ML inference → area calculation → pricing. Designed to be embedded as a 'get instant quote' widget on any landscaping company's website.

It eliminates the biggest friction point in landscaping sales: the in-person estimate. The project went through a 9-phase refactor to get the architecture right — clean service layer, proper rate limiting, separation of ML inference from business logic. It's not a demo. It's an embeddable product pipeline.

Prototype — went through 9-phase architecture refactor