Spatial intelligence for the built world

Pointcloud to CAD. Scan to BIM.

Turn 3D scans into structured building geometry. Our pipeline reconstructs floors, walls, doors, surfaces, and MEP elements from raw point-cloud data — ready to carry into CAD and BIM workflows.

  • MIT-founded team
  • Ex-Roblox & NVIDIA
  • Four years building together
Reconstructed building model with structured walls, floors, and openings
Raw point-cloud capture of the same building interior
Point cloud Structured BIM
Same capture, same viewpoint — raw XYZ points to organized building elements Drag to compare

Primary capability

From captured space to structured geometry

Limit Labs is developing pointcloud-to-CAD technology for scan-to-BIM workflows. It converts observed 3D data into organized building elements that teams can inspect and carry into downstream modeling.

  1. Input

    Point-cloud scan

    Ingest common point-cloud formats containing captured XYZ geometry and, where available, colour data.

  2. Reconstruct

    Building elements

    Identify and organize floors, walls, openings, surface information, and MEP geometry from the scan.

  3. Output

    Structured 3D geometry

    Produce coherent building geometry for review and continued work in CAD and BIM pipelines.

Current technical scope

  • Building structure reconstruction
  • Doorway workflows
  • Surface & texture analysis
  • MEP pipe identification & fitting

Additional capabilities

Generative 3D tools

Roomseed and Worldseed remain available for teams working on generated interiors and larger virtual environments.

A furnished office floorplan generated by Roomseed

Roomseed

Generate furnished interior spaces from floorplans, with room structure, object placement, and exportable 3D scenes.

Open Roomseed
An outdoor landscape generated by Worldseed

Worldseed

Create larger outdoor environments with terrain, vegetation, and architectural elements for interactive 3D use.

Open Worldseed

About Limit Labs

Spatial AI, built for real workflows

We’re builders first: MIT-founded, with experience at Roblox and NVIDIA. We’ve been working together for over four years across applied spatial computing, 3D content systems, research, and education.

We build tools that make complex spatial data easier to turn into useful, editable environments — for the people already doing the work.

The Limit Labs team

How we work

Core principles

Make something useful

We turn cutting-edge research into tools that make developers’ and artists’ work easier and more efficient.

Push the limits

We explore new approaches to spatial intelligence while staying grounded in practical customer needs.

Enable creators

We build systems that help people understand, edit, and extend 3D environments — not replace the people doing the work.

Get in touch

Have scan data you need as usable geometry?

Tell us about the capture, the formats you work in, and what you need downstream. We’ll tell you honestly whether we can help today.