/experience → clarity /
praXvalue is a deep-tech advisory firm specialising in robust decision-making in environments shaped by uncertainty, scarcity of data, and high strategic stakes.
For more than two decades, our work has focused on extracting meaningful, reliable signals from complex systems—first in institutional finance, and today in geopolitics, defence, and negotiation.
We have built a long track record in the investment industry, delivering advanced solutions for alternative signal generation, nonlinear risk modelling, and portfolio optimisation. Many of the most powerful insights in asset management arise from problems that are fundamentally NP-hard—from higher-order moment constraints in trading algorithms to structural regime detection and alpha extraction.
Our contribution has been to design convex, spectral, and semidefinite relaxations that make these problems tractable, robust, and operational at scale.
This same mathematical discipline underpins our current work in sovereign and defence-grade AI. Building on our heritage in convex optimisation, spectral geometry, and uncertainty modelling, we now develop eigenmind: an air-gapped, computationally frugal intelligence engine that collects, structures, and activates the experiential knowledge of highly advanced human operators. Through higher-order spectral analysis of proprietary semantic graphs, eigenmind surfaces relevant-but-not-obvious insights, narrative attractors, and cognitive singularities—supporting negotiators, diplomats, and military decision-makers facing adversarial pressure. Across finance and geopolitics, our mission remains the same: turn complexity into clarity by solving the right NP-hard problems with the right convex intelligence.
praXvalue stands at the intersection of quantitative science, strategic decision-making, and sovereign AI, providing the tools and advisory expertise needed to navigate a world where uncertainty is not an obstacle—but a structure to be understood and leveraged.
Grounded in decades of mathematical and AI research for robust decision systems.
Operated in complete isolation from external networks—making and suitable for classified environments and sovereign strategic decisions.
Robust systems built for real-world, high-stakes execution.
decision-intelligence engine for high-stakes environments
eigenmind is a decision-intelligence engine designed to capture, structure, and operationalise expert knowledge in environments where data is scarce, uncertain, or fragmented. Built on three decades of research in optimisation, AI, and risk modelling, it transforms tacit human expertise into reusable decision frameworks while remaining fully air-gapped and computationally frugal. To achieve this, eigenmind combines light, efficient open-source pretrained embedding models for semantic structuring with proprietary semantic-graph construction methods, and uses compact small language models for lightweight, local inference — ensuring high performance without ever compromising security. At its core, eigenmind relies on eigenvalue-based optimisation across semantic graphs to reveal structure, surface insight, and support robust decision-making.
Transforms tacit expertise into structured, reusable decision frameworks
Fully isolated and computationally frugal for sensitive environments
Reveals intrinsic structure hidden within complex systems
Built for defence, finance, and high-stakes decision domains
François Oustry is an applied mathematician and entrepreneur specialising in decision-making under uncertainty. A graduate of ENSTA Paris and Stanford, he earned a PhD in convex optimisation funded by the DGA (French Defence Procurement Agency), followed by a postdoctoral fellowship at NYU’s Courant Institute. He was then recruited as a researcher at INRIA (National Institute for Research in Digital Science and Technology) before founding his first company Raise Partner — a financial risk-management technology firm. Over more than three decades, he has designed advanced optimisation and AI systems for global financial institutions, sovereign organisations, and strategic operators working in complex geopolitical environments. He later co-founded a quant-NLP hedge fund, Suzugia, acquired by AllianceBernstein — and delivered high-reliability analytical platforms to institutional investors, sovereign wealth funds, and operational teams in sensitive, high-uncertainty contexts. He is currently CEO of praXvalue, where he develops open-source, sovereign AI capabilities — including the air-gapped intelligence engine /eigenmind/ — supporting negotiators, defence communities, and organisations facing heightened geopolitical uncertainty.
View LinkedIn Profile →