Scientific advances serving regulatory compliance
Our research explores an innovative combination of specialised multi-agent systems and techniques for translating laws and financial rules into rule engines optimised for LLMs, to verify and validate compliance across domains.
This approach couples a symbolic layer (explicit rules, operational constraints) with LLM linguistic capabilities to ensure accuracy, traceability, and adaptability to evolving regulations.
Continuous improvement of accuracy, verifiability, and explainability via a symbolic layer (rule engine) and cross‑validation with (semi-)formal proof protocols.
Creation and curation of financial and regulatory datasets aligned with normative graphs and compliance test suites to ensure relevance, accuracy, and traceability.
Optimisation of energy efficiency (LLM + symbolic logic), with distributed execution and caching of rule evaluations to reduce carbon footprint.
A specialised multi‑agent system orchestrated around a rule engine with native traceability for compliance.
Dedicated agents (extraction, normalisation, control, explanation) cooperate to encode, enforce, and audit regulatory rules.
Real‑time exchanges via interface contracts and shared schemas (ontologies, normative graphs) for reliable coordination.
Continuous updates of rules and thresholds, non‑regression tests, and human oversight to follow regulatory changes without full retraining.
Unify financial data, legal documents, and metadata into a coherent base (references, versions, scope) ready for reasoning.
Transform normative texts into structured obligations and constraints (entities, conditions, exceptions) ready for rule encoding.
Encode and evaluate rules (Prolog/Datalog/constraint graphs), check compliance, and produce proofs and counter‑examples.
Generate explanations, cited sources, and actionable audit reports (traceability, rule versions, applied decisions).
Close collaborations to co‑build normative graphs, rule corpora, and evaluation suites integrated at the core of our agents.
By uniting academic AI and regulatory expertise, we target precise, explainable, and auditable compliance that stays up to date.