Portfolio
GenAI Products

AI Interior Designer — Spatial Design Platform

Upload a floor plan → multi-agent AI performs CV segmentation, builds a wall topology graph, applies a computational math engine (golden ratio, ergonomics, photometry), and generates a 60fps 3D interior.

GitHub Repository
60fps
3D Rendering
Three.js / React Three Fiber interactive canvas
3
Style Variants
Modern / Japandi / Luxury generated in parallel
SMPTE
Ergonomics
TV viewing distance, 55cm furniture passages
60-30-10
Color System
Photometric lux + Kelvin calculation per room
The Problem

Professional interior design costs $5,000–$50,000 per project and takes weeks. Existing AI tools generate aesthetically pretty but spatially incorrect rooms that violate ergonomics, furniture clearances, and building code minimum passages.

The Solution

A Bayesian Scale Estimator converts pixels to meters without manual calibration. A Shapely CAD compiler enforces 22cm wall buffers, 55cm ergonomic passages, and door swing radius protection. Three parallel style variants (Modern, Japandi, Luxury) are generated simultaneously with PDF export.

Architecture Highlights

01

Bayesian Scale Estimator — auto-converts pixels to meters, no manual calibration needed

02

Topological wall graph with probabilistic opening detection (windows / doors)

03

Computational Math Engine: golden ratio (Φ ≈ 1.618), force-directed spring solver, photometric lux

04

Shapely CAD compiler: 22cm wall offsets, 55cm ergonomic passages, door swing protection

05

Walk-mode first-person camera with wall collision + drag-and-drop furniture placement

06

3 parallel style generation (Modern / Japandi / Luxury) with PDF export

Tech Stack

Next.js 15FastAPIPythonThree.js / React Three FiberTaskiq (distributed queues)Shapely 2.0OpenCVCloudflare R2Docker

Want something similar?

Available for Founding Engineer roles & architecture contracts.

Discuss ProjectAll Projects