Guillaume on stage at #LavalVirtual, presenting our work on #Digital #Twin technologies for #Disaster #Risk #Reduction through @simexbuilder.com.
Good to see it land with the right people!

Thank you to Anne-Sophie Clayer, Gregory Maubon and Maxime Leroy for the coordination support from #Laval2026.

Attending? Come find us.

#LavallVirtual #SimexBuilder #DigitalTwins #DisasterRiskReduction

This would not have been possible without Alain Pagani from @dfki.bsky.social, Pierre Yves Noel from Alcatel-Lucent Enterprise, Jean-Pierre Lorre and his team at @linagora.com , and all the brilliant minds who mentored and enabled us on the project.

Guillaume is on stage this morning for the Industry & Defence session at #LavalVirtual, presenting learnings on human decision-making inside digital twins - a direct outcome of our teams' work on SimexBuilder and the CORTEX2 project.

The coordination failure pattern appears in after-action reports from Hurricane Katrina, Grenfell Tower, Fukushima, and countless smaller incidents. The equipment arrived. The people arrived. The coordination did not.

What matters: Can Team A share information with Team B within required timeframe?
That is measurable. Most training programmes do not ask it because tools to measure coordination at scale have not existed.

Coming weeks: specific problems coordination failure creates.

Results:
- Grenfell: 3 services declared Major Incident independently (no mutual notification)
- Manchester Arena: "JESIP almost completely failed"
- Coordination principles "not muscle memory"
Not lack of procedures. Lack of joint training.
Most organisations measure: "We ran 2 exercises this year"

Belfast demo motivation: fixing coordination training.
#JESIP data (Joint Emergency Services Interoperability Principles):
- 33% of police commanders NEVER exercised jointly
- Only 8% train jointly every few months
Yet they are expected to coordinate seamlessly during major incidents.

What happens next: pilot exercises with real organisations, real scenarios, real participants, real feedback.
Demo day was relief. What comes next is harder.

What the demo did not prove: whether safety professionals will adopt it in their real workflows, whether the pricing model is right, whether onboarding is intuitive, whether 500 participants in a single exercise actually works at scale.
Those are different questions. They require different evidence.

What the #demo validated: architectural decisions hold up, #DTDL foundation works for exercise modelling, no-code authoring for domain experts is viable, multi-deployment is technically achievable.

Team A received the #hazmat alert at T+3 minutes.
Team B acted on it at T+11 minutes.
That 8-minute gap is where people could have been harmed.

That is measurable, evidence-based after-action review.
Not "how did everyone feel it went?"

Here is a concrete example:
The platform logs every #inject delivery, every decision point, every communication between roles.
After an exercise, the debrief is specific.

The #Technology in a #SimulationPlatform is the least interesting part.
The interesting part is what happens between people.
#Emergency #response failures are almost never caused by lack of equipment.
What breaks down is the coordination between the agencies operating that equipment.

These are architectural decisions. They determine what the platform can become, not just what it does today.

#DigitalTwin #TechnicalArchitecture #VR

The moment that made the room go quiet was not any of those five.
It was when Aura Studios showed the hands-on VR world-building preview. Seeing a training scenario rendered as a walkable environment changed the conversation from architecture to imagination.

5 things we showed at the Belfast engineering demo. No slide decks.
No-code exercise authoring.
Semantic foundation built on Digital Twin metamodels.
Local generative AI for context-aware interactions.
Multi-deployment architecture: cloud, on-premise, offline edge. Interoperability by design.

The 10% tradeoff? Our scenario design tool does not handle participant data.
The sensitive information—who performed well, where the coordination gaps are—stays on European infrastructure.

Deliberate. Enforced at the architecture level, not just policy.

European tech stack in practice: EU-hosted infrastructure, European-owned providers, GDPR-native architecture, no US data transfers as default, no risk of unlawful US CLOUD Act interference.

We made a deliberate decision: >90% of SimexBuilder's v0.1 technology stack would be European owned.
Emergency simulation data is sensitive.
Participant performance records.
Organisational vulnerability assessments.
Scenario details describing weaknesses in infrastructure and response protocols.

Tabletop exercises max out at 30-50 people. If your organisation has 500 staff, you do the same exercise 17 times. Or train 30 and hope the other 470 absorb it through osmosis.

Most corporate organisations keep emergency training exercises in a tabletop format:
sitting in a conference room talking about what they would do.

Then face the reality of staff freezing when it actually happens.

Staff who understand technology, linked data, and twin semantics.
Why does that matter for a simulation exercise platform?

Because SimexBuilder's simulation exercise is a digital twin of a training scenario. Same structural concepts used in building management, applied to emergency training.

Belfast has something most European cities are still trying to build.
Most people outside Northern Ireland do not know about it.
The Digital Catapult Digital Twin Centre is not a co-working space with a fancy name. It is a facility where digital twin technology meets industrial application.

@linagora.com @actioncontrelafaim.bsky.social @ec.europa.eu @dfki.bsky.social @cortex2.bsky.social

@linagora.com @actioncontrelafaim.bsky.social @ec.europa.eu @dfki.bsky.social @cortex2.bsky.social

If anyone works with #digital #twin #modelling and is curious about applying it to #disaster #risk reduction #simulations, we are happy to discuss.

The Belfast demo was our first external validation.
Architects, engineers, digital twin specialists asked hard questions.

Our specifications survived.

That coherence? It took 18 months of doubt as the price. But it was worth it.

Builder #UI, execution #gateway, participant #apps.
Same specification, same data model.
It works coherently because the data interfaces were defined before the code.

The tension: it is slow. Colleagues asked "when can I see it?"
We said "not yet."
For months.

You wonder if you have over-specified.
If the architecture is too rigid. If you should have just started building.
Then one day, everything runs together.