Starting with a blank sheet, we slashed Hover's unit cost by 75% in the
                first two years.
Core features include interactive photogrammetry for solving camera networks, CV
                surface texturing and ML-powered feature recognition. We started with a core suite of parametric tools,
                taking advantage of deep knowledge of our user's process in other modeling apps.
We used
                property data, local code, and CV to predict gutter locations and paths while a user draws. Automation
                only makes suggestions when they're likely to be helpful.