Spatial Engine AI

The First AI That Sees Light Like a Physicist. Deterministic Physics + Generative Vision.

Comment

Spatial Engine is a next-generation AI agent designed to bridge the gap between digital intelligence and physical environments. Built on the Gemini 3 Pro model, it solves a critical problem in modern living: inefficient lighting and energy waste.

Using multimodal spatial-temporal reasoning, the agent analyzes live video feeds to understand the physics of a room—identifying window positions, furniture layouts, and shadow zones. By executing autonomous Python code, Spatial Engine calculates the optimal placement for light sources and solar panels, providing users with a "Lux Optimization Map."

This is not just an analyzer; it is an action-oriented engine that helps reduce residential energy consumption by up to 40% and improves human well-being through scientifically optimized environments. Our mission is to make every photon count in the "Action Era".

Inspiration

We realized that current AI models are great at chatting but terrible at engineering precision. When it comes to lighting design, "hallucinating" a lux level is dangerous. We wanted to build an agent that sees like a designer but calculates like a physicist.

What it does

Spatial Engine AI is a multimodal autonomous agent. It audits rooms via photos, calculates exact lighting deficits using the Inverse Square Law ($E=I/d^2$), and automates the procurement process.

Key features:

  1. Physics Core: Deterministic math engine (not LLM guesses) for Lux and Energy calculations. Includes ISO/SanPiN health compliance checks.

  2. Market Intelligence: Real-time search for products (verifying "dimmable" specs) and electricity rates to calculate exact ROI and Payback Period.

  3. Vision System: Uses Gemini 3 Vision to decompose room geometry, detect materials, and estimate scale.

  4. Resilience: Works offline via a robust Fallback Mode.

How we built it

The brain is Gemini 3 Pro (accessed via Google GenAI SDK). The core logic is written in Python, utilizing a custom PhysicsEngine and MarketAgent. We use pypdf for reading datasheets and matplotlib (in progress) for visualization. The frontend is built with React, Tailwind, and Framer Motion to visualize the agent's "thought process."

Built With

Share this project:

Updates