Problem
Modeling light gauge steel framing walls by hand — studs, tracks, openings, tight tolerances — is slow, and it repeats the same risk on every project: one manual pass, one chance to introduce an error the shop floor inherits.
Selected Work
Projects.
Three examples of AI-driven engineering automation, chosen for what matters in the boardroom, not just the CAD session: can it move faster, can it be trusted with sensitive data, and does it actually produce manufacturing-ready output. Not construction-specific — the same operating model applies across engineering and manufacturing domains.
Case Study 01
Automated LGSF Wall-Panel Generation
Manufacturing-ready framing models generated from governed rules — not redrawn by hand on every project.
Approach
Wall systems are generated from governed templates and knowledge-driven rules inside CATIA and 3DEXPERIENCE, so every wall, opening, and framing member comes from the rule set instead of being placed by hand, project after project.
Outcome
A structured model built to LOD 400 — manufacturing detail, not a design sketch that gets rebuilt downstream. For the business, that's engineering hours converted into throughput: the same team carries more projects without the modeling bottleneck growing headcount.
Case Study 02
Secure Data. Smart AI. Protected Clients.
The question every executive asks before approving AI: where does our data actually go? Here, the answer is: nowhere it shouldn't.
Problem
Most organizations' information is already scattered — network drives, SharePoint, email, ERP, PLM, partner files — with no single owner and no clear current version. Layer AI on top of that by uploading raw documents to a model, and exposure multiplies instead of shrinking.
Approach
Documents stay governed inside 3DEXPERIENCE. AI logic reads them inside that secure environment, extracts only the specific data needed, formats it as structured JSON text, and sends only that — never the source file — out for processing.
Outcome
No raw document ever leaves the building. Only minimal, structured text does. That's the difference between "we use AI" and being able to tell legal and the board exactly what AI can and cannot see.
Case Study 03
Mass Timber Structure, AI-Powered
The current engagement: an AI agent reads the spec, builds the CATIA model, publishes it, and evaluates its own work — with a human reviewing the output, not producing it by hand.
Problem
Taking a structure from an approved engineering spec to a manufacturing-ready CATIA model is normally a manual, project-by-project effort — slow to start, and slower every time the spec changes.
Approach
An AI agent reads the current spec, drives CATIA branch and geometry generation directly, publishes the result, and is then asked to evaluate its own output — for example, whether column sizing should optimize by floor — before a human signs off.
Outcome
A revision that used to take a modeling effort now takes minutes of agent work. This is what AI actually doing engineering work, not just assisting with it, looks like in production — and it's the same operating model behind every case study on this page.
Have a workflow like one of these?
It's worth a conversation.
If your data, your models, or your AI ambitions look like any of the three case studies above, send the specific problem — a modeling bottleneck, a data-exposure risk, or an AI initiative stuck at the proof-of-concept stage — and I will tell you plainly whether it is a strong candidate for a proof of concept.