About
Built by architects and engineers.
Half practice, half research. Ex-practitioners from architecture and engineering firms, working alongside researchers published at ICML and NeurIPS in computer vision, AI agents, and complex document retrieval. The agent logic is written by people who have done these reviews themselves, not by AI generalists.
120+
Checks in the library
Architectural · Structural · MEP · Civil
20 min
To review a full set
Every sheet gets the same pass, at any project size
9 in 10
Findings confirmed
Triaged and fixed by client engineers, not noise
01 · Why we exist
A drawing set is a legal document that nobody has time to read properly. The errors that survive review don't stay on paper: they become RFIs, change orders, and rework.
Reviewing a five-hundred-page package against the fire report, the schedules and the code is slow, unglamorous, and the first thing compressed when a deadline moves. It is also the work that decides what the project costs.
We started Structured AI because we had done that review ourselves, at firms where the consequences of missing something were ours to carry. So we built agents that read a set the way a reviewer does, cross-referencing sheets, schedules and standards, and return every finding cited to the page it came from.
02 · How we work
Practitioners write the logic
Every check is specified by someone who has produced and stamped the drawing it inspects. The agent reasons the way a reviewer does, against the fire report, the schedule, the code, not against a keyword list.
Every finding is cited
A result is only useful if it can be verified in seconds. Each finding lands on its exact sheet and location with the reasoning attached, so an engineer can confirm or dismiss it without reopening the set.
Your drawings stay your advantage
Your project data never enters model training, and every organisation runs in its own isolated environment, pilots included.
Read the security model →03 · Where we come from
Practice
Our architects and engineers practised at
They produced and stamped the drawings the checks now inspect.
Research
The research behind the product was done at
Published at ICML and NeurIPS.
Backing
Backed at seed and pre-seed by
Y Combinator, 20VC, Cherry, Sequoia Scout and others.
04 · In practice
“By partnering with Structured AI, we got a tool custom built for how we work, and our engineers genuinely like using it. It’s solving complicated problems, the kind where the system has to analyze and interpret the documents the way an engineer would.”
Robert Ioanna Executive Principal, Chief Technical Officer & Board Member, Syska Hennessy Group05 · Join us
We’re hiring engineers, researchers and practitioners.
Small team, real ownership, and a problem that only gets solved by people who understand both the drawings and the models.
Put practitioners’ AI on your next set.
Built by people who’ve stamped drawings. See what it catches on yours.
Book a Demo