Grihafy
Grihafy reads a home's floor plan and explains how its rooms line up with Vastu, the traditional Indian principles for orienting a home.
Problem
Real floor plans are messy: some have no clear compass, some span several floors, and some rooms cross more than one zone. An AI model can recognise rooms, but every conclusion depends on which direction each room faces.
Decision
I founded Grihafy and own its product, engineering and deployment. I rebuilt the analysis as three stages: compass detection, room detection with normalised geometry, then the Vastu analysis. Direction assignment moved out of the model into deterministic TypeScript that works from the room geometry. When a plan is ambiguous, the user confirms north, or the direction is marked unknown instead of guessed.
Outcome
Analyses run in the background with idempotency checks and rate limiting, stream their progress to the browser, and are scored against a canonical Vastu knowledge base. Grihafy is live at grihafy.com.