QSELive · Level 1Level 2 · Coming soon

Qode Spatial Engine

QSE converts OpenUSD scenes into named, versioned, queryable spatial graph assets and compressed GLB files for rendering. It is the intelligence extraction engine of the Qode Scene Intelligence stack, enabling AI agents to navigate and reason over physical environments without reconstructing raw geometry.

The Engine · Engineering
Pipeline output
USDQSG + GLB
two outputs · one USD source
QSE pipeline diagram showing Level 1, Level 2, and Level 3 evolution

Pipeline evolution

Three stages. One engine.

QSE Level 1 is live and shipping. Level 2 automates graph extraction. Level 3 requires NVIDIA GPUs to realize semantic depth at property portfolio scale.

Now
Live · Shipping
Level 1 · CPU
  • Parses OpenUSD scenes using usd-core (Pixar OpenUSD Python bindings)
  • Extracts named spatial entities, typed edges, and relationship metadata
  • Produces a typed spatial graph (QSG) for intelligence and query
  • Produces a Draco-compressed GLB for QSV rendering
  • Deterministic geometry processing. No ML, no CUDA dependency
  • Five named QSG assets in production registry
Near Term
Automated
Level 2 · CPU
  • Automated graph extraction from USD scenes, eliminating hand-authoring
  • Any USD asset becomes a QSG at pipeline speed
  • QSG Graph Hub: formal versioned registry with diff, changelog, state commit
  • Management Console: Human Super Actor governance UI
  • Scalable partner onboarding across Travel and Hospitality
  • NVIDIA Omniverse platform integration for USD capture vendors
Long Term
NVIDIA Powered
Level 3 · GPU
  • GPU-accelerated semantic parsing of USD scenes at property portfolio scale
  • Richer vector embeddings per spatial entity, depth impossible at CPU scale
  • Occupancy Data layer: real-time occupancy per node from sensors and simulation
  • Flow Data layer: pedestrian flow rates per edge via NVIDIA Isaac Sim
  • QQP (QSG + Occupancy Data) = dynamic operational intelligence
  • 5 agents to 5,000 agents · same QSG, NVIDIA GPU encoded once

Architecture and technical details

The full QSE technical architecture, including pipeline implementation, stack dependencies, and the roadmap to GPU-accelerated Level 3, is documented in our white paper.

Read the white paper

Where QSE sits in the stack

QSE produces it. QSG stores it. QSV renders it. QQP queries it.

QSE is the entry point of the Scene Intelligence stack. Every QSG asset, every QSV rendering, and every QQP query begins with a QSE pipeline run on an OpenUSD source. No QSE, no stack.