Merlin Intelligence holds that value resides primarily in people — in their experience, their discernment, and their capacity to act together. This is who we are.
Marion works at the intersection of applied mathematics, artificial intelligence, and cognitive science. She holds an engineering degree in Applied Mathematics and a PhD in Artificial Intelligence, complemented by academic training in cognitive sciences.
Her work is concerned with a question that sits at the heart of Merlin Intelligence: how does knowledge acquire structure? This leads her to knowledge representation — graphs, ontologies, the formal architectures through which meaning becomes computable — and to the modern systems that operate within those architectures: large language models, retrieval-augmented generation, agentic pipelines. She is interested in where these two traditions meet, and where they diverge.
A particular focus of her research is Ontology Learning — the methods by which conceptual structure can be extracted, induced, and refined directly from data and language, rather than authored by hand. It is this line of work that allows cognitive assets to be built and enriched at the pace experiential knowledge actually demands.
She pays particular attention to questions of explainability — to the conditions under which an AI system can be understood, audited, and held to account. This is not a constraint she works around. It is a design principle.
Her career spans academic research, including publications through public and private collaborations, and applied industry work supporting organisations in designing and delivering AI-driven projects. She brings both the rigour of the researcher and the practical judgment of the practitioner.
François is an applied mathematician and entrepreneur. His formation is in optimisation — convex, robust, and spectral — trained at ENSTA Paris, with a PhD funded by the DGA (French Defence Procurement Agency) and a postdoctoral fellowship at NYU's Courant Institute. He was then recruited as a researcher at INRIA before turning toward practice.
Over three decades, he has designed analytical systems for institutions operating in conditions of genuine uncertainty: global financial institutions, sovereign organisations, strategic operators navigating complex geopolitical environments. He founded Raise Partner, a financial risk-management technology firm, and later co-founded Suzugia, a quantitative and NLP-driven hedge fund acquired by AllianceBernstein.
What connects these engagements is a consistent preoccupation: decision-making when the model is incomplete, the environment is adversarial, and the cost of error is asymmetric. The answer is never more data. It is better structure — tighter representations, more honest uncertainty, and the capacity to act on weak signals before they become legible to everyone.
He is currently CEO of praXvalue. Eigenmind, the open-source intelligence engine he originated, has since grown into a fully independent, open-source resource — and a key capability within Merlin Intelligence — supporting negotiators, defence communities, and organisations facing heightened geopolitical uncertainty.
He is also in charge of the "AI Do It Yourself" practice at Orson.ai, a strategic advisory firm — extending the same conviction that drives his work at Merlin Intelligence: that organisations gain the most when they are equipped to build and own their own intelligence, rather than rent it.