From the LABFor the floor
Two applied research axes: sovereign software engineering and industrial human factors.
Open research questions, tested against industrial reality. Technical barriers that open the way to new approaches.
The challenge?Rethinking industrial intelligence.
Explore the researchOur starting point
Why do industrial tools still lag behind the technologies people use every day in their personal lives?
This question is where everything began. Our R&D hub exists to address it — methodically, rigorously and over the long term.
The finding
Industry invests and digitalises. Yet the expected gains still struggle to materialise on the floor. Tools were designed for processes — rarely for the people who bring them to life.
The symptom
KPIs produced for meetings can be so difficult to calibrate that they are already obsolete when analysed. Meanwhile, part of the operational reality remains invisible.
The legacy
Older applications extended layer after layer — powerful on paper, yet increasingly heavy, fragmented and rigid in use.
The opportunity
This underused human potential represents substantial value for companies and European industrial competitiveness. It is also an opportunity to design an industry that speaks the language of a new generation.
Our research begins in this gap — between systems, usage and the reality of the floor.
Our convictions
Industry was largely left behind by the digital revolution
Many industrial environments still rely on ageing architectures, sometimes modernised only at the surface.
People are still expected to adapt to tools — rarely the other way round.
Context is the missing data
Industrial systems record the information required for their own purpose. ERP represents management, CMMS maintenance, MES production, and quality systems inspections and deviations.
The full context of a situation does not reside in any one of these systems. It lies in the relationships between their data, the actions performed and observations from the floor.
The challenge is not only to bring more data together, but to preserve the relationships that give it meaning.Sovereignty is a condition of control
Production data reveals processes, performance, vulnerabilities and part of a site's competitive advantage.
For critical industries, control cannot be separated from architectural, hosting and jurisdictional choices.
Industrial digital systems must also be efficient
Data movement, compute, storage and network infrastructure all have a footprint.
Designing leaner systems without compromising reliability is a research problem in its own right.
It is a responsibility, not a claim.Phase 0 · Research & Technology
Research starts upstream.
The most structural decisions are often made before a project officially begins. While needs, uses and architecture remain open, assumptions can still be challenged and early directions tested against the reality of the system.
Reduce uncertainty
Observe the system as a whole. Separate established knowledge from assumptions. Identify the constraints, risks and barriers that genuinely shape the project.
- Observe
- Frame
- Formulate
- Experiment
Drive architectural convergence
Compare scenarios with the evidence obtained. Reject assumptions that do not withstand experimentation and build a path consistent with the maturity actually demonstrated.
- Measure
- Compare
- Decide
- Orient
An early decision should not merely be plausible. It should be open to challenge, experimentation and defence.
Technology foundation · R&D Hub
PULSAR — The foundation that makes exploration possible.
Industrial real-time middleware written in C++ for experimenting with distributed systems, state consistency, sovereignty and embedded intelligence.
TRL 9 · Deployed
PULSAR
Industrial real-time middleware
- Reactive state synchronisation and a coherent source of truth.
- Cryptography and data control integrated into the architecture.
- C++ core and dedicated domain language.
TRL 3 · In development
PULSAR MAINT LM
Industrial maintenance AI
- Model specialised in industrial diagnosis and procedures.
- Reasoning augmented with deterministic tools.
- Logical and safety verification with traceability.
TRL 3 · In development
PULSAR AI EDGE
Autonomy on the floor
- Distillation into lightweight models for isolated sites.
- Local and offline operation.
- Reduced network and cloud dependency.
Axis 01 · Sovereign software engineering
Sovereign software engineering
Exploring architectures that make industrial systems more responsive, verifiable, autonomous and efficient.
Software sovereignty in distributed environments
How can industrial data remain under technical and legal control while maintaining real-time performance across a multi-site network?
A matter of strategic autonomy and operational continuity for sensitive environments.
Computational efficiency without loss of fidelity
How can data flows in a distributed system be substantially reduced without degrading information consistency?
Reducing infrastructure cost, energy footprint and network dependency.
Low-footprint embedded AI architectures for constrained real-time systems.
LLM distillation for offline deployment
How can a model be compressed while retaining specialised industrial diagnostic capabilities?
A prerequisite for deploying AI directly on site without permanent cloud dependency.
Distillation models for edge deployment in offline industrial environments.
Hybrid deterministic–probabilistic reasoning
How can probabilistic model reasoning be combined with deterministic standards, business rules and physical models?
Every recommendation must remain traceable, verifiable and compatible with critical-environment requirements.
Hybrid reasoning systems for reliable industrial AI.
Axis 02 · Industrial human factors
Industrial human factors
Understanding how industrial actors produce, transmit and transform knowledge through real work.
The human as sensor
How can natural observations and tacit knowledge become usable information without adding reporting burden?
Systems can only analyse what they already know how to observe. A decisive part of reality remains carried by people.
Capturing tacit knowledge through adaptive interfaces in industrial settings.
Industrial Threading — preserving context
How can relationships between management, production, maintenance, quality and real activity representations of the same situation be preserved?
Each system accurately describes one part of reality. Intelligence also lies in the relationships between these viewpoints.
A model of the industrial thread linking action, environment, intent and consequence.
Sustained engagement in industrial systems
How can progression and recognition mechanisms be designed without habituation, manipulation or displaced motivation?
Adoption cannot be imposed: it depends on perceived usefulness, autonomy and recognition of real work.
Observable skills and execution data
How can a verifiable representation of skills be built from real situations without reducing people to a metric?
A topic at the intersection of knowledge, ethics, evaluation and work organisation.
Skills frameworks based on contextualised evidence of execution.
Open work
Questions to investigate across disciplines.
Some of this work may lead to academic exchanges, experimental settings, doctoral research or collaborative programmes.
- Distributed systems
- Reliable AI
- Edge computing
- Human factors
- Cognitive science
- Industrial ergonomics