Praedon, a spin-off from CERMICS and CIL, awarded the Grand Prix at the iLab 2026 competition

The start-up Praedon, a spin-off from CERMICS and the Co-Innovation Lab, is one of the 10 Grand Prix winners of the iLab 2026 competition, a competition for innovative deep-tech start-ups organised by the Ministry of Higher Education and Research.

The Grand Prix category recognises projects “deemed particularly promising, with a particular focus on those with a significant impact in terms of sustainable development and societal benefits”.

Praedon creates AI foundation models that combine deep learning and optimisation to maximise business KPIs whilst ensuring compliance with complex industrial constraints

Today’s most critical industries – from global supply chains and transport to energy networks – rely on incredibly complex, crucial decisions. To optimise these networks, companies have traditionally had to choose between two imperfect solutions: rigid mathematical solvers that struggle to adapt to real-world uncertainties, or modern Artificial Intelligence which, whilst highly predictive and flexible, is probabilistic and cannot guarantee that its decisions are physically or operationally feasible.

Praedon is exploring a new frontier in enterprise software to solve this problem. It designs ‘Industrial World Models’ – a revolutionary AI architecture that merges the predictive intelligence of deep learning with the mathematical guarantees of Operational Research. This AI does not merely learn and predict events from big data, but focuses on what really matters: decisions that comply with the physical and business rules of the real world.

This technology has a profound impact and has already proven its worth in practice. By optimising complex decisions, the solution generates millions of euros in operational savings for industry leaders. For example, in automotive logistics, our models have optimised vast transport networks, saving over one million euros a year whilst simultaneously reducing CO2 emissions by 10 per cent. Similar successes have been achieved in use cases relating to energy trading and production planning. Furthermore, as our models use structured architectures, they require significantly less computing power and energy than standard large language models, thereby contributing directly to a more sustainable technological ecosystem.

Through this project, Praedon is commercialising a scientifically validated breakthrough, transforming it into a highly scalable and easily integrable software platform. By facilitating resilient and actionable decision-making in our physical world, we are empowering businesses and governments to achieve high levels of efficiency, sustainability and strategic autonomy.