Marius Dima
Building AI systems that explain themselves. Neural-symbolic architectures for critical systems where trust isn't optional.
Background
20 years: extreme industrial automation → manufacturing automotive process & conversational AI → neural-symbolic AI
If it can't explain itself, it shouldn't run your factory. Or your democracy.
3 published patents. Speaking: MIT, AI Summit London, How to Web
Currently: Qriton (trustable AI) + iGov.ro (tools for transparent governance)
Core Focus
Neural-symbolic AI architectures that merge deep learning with logical reasoning
EU AI Act compliance frameworks for high-risk systems
Explainable decision-making where transparency is non-negotiable
Energy-based models for anomaly detection in critical infrastructure
Projects
QRITON Active
Neural-symbolic AI platform / EU AI Act compliance / Explainable decision-making for critical systems
@QRITON/HOPFIELD-ANOMALY Production
NPM package / Unsupervised anomaly detection / Energy-based models
IGOV.RO Active
Civic tech / Open governance / Citizen participation platform / TDI member
AIRADIO.HOST Research
Content automation / Transparency experiments / Algorithmic broadcasting
CONFYCHAT Research
Secure communications / Privacy-preserving protocols / E2EE research
BADTOK Archive
Algorithmic accountability / Social platform transparency / Content moderation research
Book
LUT ABSOLUT
A novel exploring the absolute threshold between systems and chaos. Available in Romanian from Librăria Delfin.
View Book →