Applied Research Services
Applied research solutions & Tangible projects
Sometimes, research finds.
Efficient AI Solutions
Symbolic AI with Formally Constrained Machine Learning
Integrating formal systems with probabilistic inference engines to enable provably correct neural architectures. Our framework combines context dependent languages, curry-howard isomorphisms and a lot of types to constrain ML training within formally verified specifications. Applications include more efficient theorem proving, correct-by-construction code generation, and decidable constraint satisfaction to help decision-makers.
a sight of the productive future
Macroeconomics prediction Models
Applying learnt rules to create synthetic world models and help you see the edge of tommorow. Outlining social trends, long-term market behavior and probable collapses of overarching structures. Shaping a model to define the future.
Personne ne sait ce qu’il se passe aujourd’hui parce que personne ne veut qu’il se passe quelque-chose. En réalité on ne sait jamais ce qu’il se passe, on sait simplement ce qu’on veut qu’il se passe et c’est comme ça que les choses arrivent.
Analog Neuromorphic Hardware
Interaction Net Processors for Biologically-Inspired Computation
Designing analog VLSI circuits that implement interaction combinators and symmetric interaction calculus at the hardware level. Our architecture enables massively parallel graph rewriting with constant-time β-reduction through asynchronous, event-driven computation. The happy fly™ project aims to simulate 10e16 Drosophila neurons (complete connectome) using interaction net substrates, achieving neuromorphic efficiency while maintaining formal correctness guarantees.
Art is distinguished from Nature, as doing (facere) is distinguished from acting or working generally (agere), and as the product or result of the former is distinguished as work (opus) from the working (effectus) of the latter.