STRUCTURAL INTELLIGENCE · AI · HKUST

from concept to code-compliant structural design in seconds

saXes.Ai is an AI design copilot that runs the full structural workflow — model, analysis, and code-compliant design in seconds — with the engineer in command.

Peer-reviewed benchmarks ·F1 0.928 recognition ·code-compliant in 18 s ·engineer-in-the-loop
PRODUCT · HOW IT WORKS

A design copilot that runs the whole structural workflow.

A specialised AI agent — a “Copilot” for structural engineering — turns sketches, drawings, text or voice into code-compliant design, with the engineer in the loop at every step.

INPUT
Sketch · drawing · text · voice
RECOGNISE
Structural model
ANALYSE
Physics-based analysis
DESIGN
Code-compliant design
OUTPUT · 18 s
Drawings + reports
Recognised digital structural model
RECOGNISE
Structural model

Multi-modal input becomes a clean, editable structural model — nodes, members and loads.

Bending-moment diagram
ANALYSE
Structural analysis

Proprietary finite-element solvers compute real physics — not an LLM guess.

Reinforced-concrete rebar design
DESIGN
Code-compliant design

Rule-based checking against design codes produces compliant, ready-to-issue output.

Just say what you want
CONVERSATIONAL
“Add two storeys on top of the structure.”
Re-modelled, analysed & code-checked in 18 s. ✓ compliant
a “Siri / Copilot” for structural engineering
EN · 普通话 · 廣東話
The engineer stays in the loop
Review the generated design
Modify in plain language if needed
Decide & approve the output
Every approval and edit feeds a fast feedback loop — the copilot re-runs analysis and design until you sign off.
VALIDATION · BENCHMARKS

The LLM never computes results — physics and codes do.

A hybrid architecture is what makes the output trustworthy. Language models read intent; proprietary FEM solvers and rule-based code checking guarantee correctness.

0.928
Structural recognition
F1 · 18.05 s · 1.14 steps
0.898
Conversational modification
F1 · 12.35 s
18 s
Intent → compliant design
model · analyse · design
Physics-derived results
Proprietary nonlinear-dynamics finite-element solvers handle analysis.
§
Rule-based code checking
Design follows state provisions and design codes — checked, not guessed.
Peer-reviewed
Benchmarks published and under review — validation, not marketing.
TEAM

Structural engineers and AI researchers, on four continents.

Prof. Elias Dimitrakopoulos
Prof. Elias Dimitrakopoulos
FOUNDER · CSO
25+ yrs structural engineering · HKUST · Cambridge · European market access · consulting & structural-design experience.
Dr. Cheng Ning Loong
Dr. Cheng Ning Loong
CO-FOUNDER · CTO
Architect & lead author of the validated AI-agentic workflow · delivery lead.
×3
1 engineer + 2 developers
HKUST
Recognition · RC design automation · visualisation.
Academic representatives, four continents
PE · CEng/IStructE · ΤΕΕ · CPEng
Senior academics across the USA, Europe, Asia & Australia/NZ — registration & local consulting lead.
VISION

To lead the paradigm shift in structural engineering.

AI accelerates and automates design, code-compliance guarantees safety, and the human engineer remains in command.

Eliminate tedious workflows
No more re-typing drawings into FE software or checking codes by hand.
Multiply engineering capacity
Every engineer ships more compliant designs in less time.
Empower high-value judgment
Engineers spend their time on decisions, not data entry.
EARLY ACCESS

Join the waitlist.

Be among the first engineering teams to put the copilot to work. We’ll reach out as access opens.

FOUNDER · CSO
[email protected]
CO-FOUNDER · CTO
[email protected]