PROSuiteResearch · Neuro-symbolic AI
AI tools that reason, to run the business
The VAILS server and its VVL language bring together language models, which understand text, and symbolic reasoning, which applies rules exactly and verifiably. Our research has produced high-level tools, each designed for a family of management decisions.
Jev model
A fast decision model whose confidence is measured, and our benchmark comparing a language model alone with neuro-symbolic modelling.
Read more →Universal resolver
A natural-language question is classified, then handed to the right engine: calculation, optimization, logic, expertise or extraction.
Read more →Optimization & planning
Staff schedules, routes, cutting, assignments: the best solution under constraints, and the explanation when there is none.
Read more →Expert system
Describing a domain and its rules in plain language is enough to obtain an expert system: ontology, reference profiles and fuzzy rules, ready to decide.
Read more →Ontologies & business rules
The company's categories, rules and exceptions, written once and applied exactly to documents and data.
Read more →Rule learning
Finding, from labelled examples, the readable rule that explains them: AI that learns without becoming a black box.
Read more →A common principle: the model proposes, the symbolic layer decides
All these tools follow the same rule. The language model is used to understand: read a document, interpret a question, propose a structure. The decision is handed to a symbolic engine: a constraint solver, a logic engine, an ontology, a set of learned rules. What is decided can therefore be verified, explained and replayed.
- Understand
- Model
- Decide
- Verify and trace
These tools are instructions of the VVL language. They combine in a single program: an ontology describes the data, a learned rule classifies it, an optimization problem derives a schedule from it, and the resolver answers questions asked in natural language.
Apply this work to your processes?
The diagnosis starts from how you actually work and identifies the decisions that can be entrusted to AI.