Agents cannot read spatial data
Language models do not understand coordinates, geometry or layers, so answers about places are guesses.
AI agents that understand maps, models and drawings
We build AI that works with spatial and engineering data, not just text: assistants that query GIS data with real tools, knowledge graphs that give agents connected context, and recognition of drawings and scans.
The problem
Why the data does not work together
Too much context, not the right context
An agent searching a whole enterprise graph gets lost. It needs the slice relevant to the place and decision at hand.
Drawings locked in images
Floor plans and engineering drawings exist only as scans, photos or PDFs that software cannot use.
Unlabeled 3D data
Reconstructed meshes and point clouds are one surface, with no idea which part is which object.
What we build
Software we build for this
Delivered as a module inside your product or alongside your enterprise stack, on open standards and without lock-in.
Spatial AI assistant
An MCP server and conversational interface over your GIS data and analytics, built into your product.
Knowledge graphs for twins and agents
Ontology and data in a graph database, with reasoning to infer links between assets and places.
ML segmentation of scans and meshes
Semantic segmentation and classification of reconstructed 3D data.
Drawing to CAD
Scans and images of engineering drawings converted into SVG and DXF vectors with VectraXD.
Floor plan and drawing recognition
Images of floor plans and engineering drawings turned into structured data: rooms, walls, symbols and text. Part of VectraXD.
Case studies
Where we have delivered this
Telecom
Knowledge Graph for Telecom and Real Estate Assets
A knowledge graph on AWS Neptune linking telecom network assets with the real estate they sit on, built on the RealEstateCore ontology with reasoning.
Read case study
GIS / Urban planning
MCP Server and Conversational Interface for a GIS Platform
Existing GIS data and analytics exposed as an MCP server, with a conversational interface in the product.
Read case study
Reality capture / Survey
3D Mesh Reconstruction with ML Segmentation and Adaptive Point Clouds
Segmented, semantically classified 3D meshes reconstructed from images, using ML segmentation and adaptive point clouds.
Read case study
Manufacturing
VectraXD™ Vectorization: Engineering Drawing Images to SVG and DXF
The vectorization module of our VectraXD™ platform, a SaaS tool that converts photos, scans and PDFs of 2D engineering drawings into SVG and DXF.
Read case study
Capabilities
The engineering behind it
Technologies and standards
Open standards, no lock-in
Prolincur is a member of NVIDIA Inception and Microsoft for Startups Founders Hub.
Industries
Where this applies
How we deliver
From first call to long-term support
- 01
Discovery call
30 minutes with an engineer, under NDA if you need one.
- 02
Technical discovery
1 to 3 weeks: requirements workshop, code and data review, and a proof of concept for the risky parts.
- 03
Proposal
Phased plan, team, estimate and engagement model.
- 04
Delivery
Two to three week iterations, with a demo every sprint.
FAQ
Common questions
Which AI models do you work with?
We build around the model your team prefers and expose your data through MCP, so the assistant uses real tools instead of guessing.
Is our data sent to a public AI service?
Only if you choose that. We can deploy inside your cloud or on-prem, with the model and data under your control.
Is floor plan recognition available now?
It is in development as part of VectraXD. Drawing vectorization to SVG and DXF is available today.
Need this in your product?
Tell us what you are building. You will talk to a software engineer from the first call.