Research scientist · Consultant

Hugolin Bergier

Logic-based AI for program understanding

When AI explains software, its statements about the code should be verifiable. As Research Scientist at Phase Change Software since 2017, I lead the design of a formal ontology of program semantics, designed the derivation logic that reasons over it, and am now connecting that layer to LLM agents. Through Analogia Entis, my independent practice, I consult on formal ontologies and knowledge graphs, hybrid symbolic and LLM architecture, and verification in Lean 4.

Open to research-scientist roles · Remote, US and EU time zones · contact@analogia-entis.com

Work with me Read the CV

3,182

COBOL modules and 3,332 copybooks of a Fortune 100 insurer, ingested into one knowledge graph

Lean 4

machine-verified counterexamples that identified soundness issues across two revisions of a client's framework

US + EP

granted patent family (three grants) on machine-based instruction editing

PhD

mathematics and formal logic, Sorbonne; Visiting Professor of Computer Science, Regis University (2026 to 2027)

Selected Work

Industrial and Consulting Work

Items 01 to 04 and 06 are my work at Phase Change Software on a deterministic code-understanding platform. Item 05 was a consulting engagement through Analogia Entis.

A formal ontology of program semantics

Lead designer of the platform's formal ontology for program semantics: 19 entity types and 19 relation types, organized in four layers (computational, context, business, and bridge).

Ontology engineeringKnowledge representationProgram semantics

Applied to a Fortune 100 COBOL estate

The ontology was applied to the COBOL estate of a Fortune 100 US insurer: 3,182 modules and 3,332 copybooks ingested into a knowledge graph supporting structural and semantic queries.

COBOLKnowledge graphs

Derivation logic over the graph, portable to SQL

I designed schema-level derivation logic, written as TypeQL functions over TypeDB, that computes functional classifications and aggregations of program structure from the schema. The derivation logic has been shown portable across a graph database and a SQL engine, which decouples the ontology from any single store.

TypeDB / TypeQLAutomated reasoningSQL

Patent family on machine-based instruction editing

Co-inventor on the granted patent family Inductive Equivalence in Machine-Based Instruction Editing: US 11,157,250 B2 (2021), US 11,720,334 B2 (2023), and EP 17825321.7 (2025).

Program analysisPatents

Machine-checked review of a complexity-theory framework

For a consulting client, a full formalization in Lean 4 (with Mathlib) of a novel complexity-theory framework. The machine-verified counterexamples identified soundness issues across two major revisions of the paper.

Lean 4MathlibFormal verification

Connecting the symbolic layer to LLM agents

Research prototypes and reasoning engines in Prolog, Haskell, and Clojure. Current work integrates the symbolic layer with LLM-based agents, using Claude Code, MCP servers, and multi-agent harnesses.

Neurosymbolic AIMCPPrologHaskell

Publications

“Tokens Are Not Symbols: Why Large Language Models Lack Rationality.” Human and Artificial Rationalities (HAR 2026), Lecture Notes in Computer Science, Springer. In press; presented in Paris, September 2026.
Eabrasu, Bauman, Bergier, and Murtinu. “Agency in the Loop: The AI Revolution and the Philosophy of Management.” Philosophy of Management, Springer, 2026. DOI
“Can Formal Logic Make Pure Chance Intelligible? Ladrière on the Eschatological Horizon of Reason.” Logica Universalis, Springer Nature, 2025. DOI
“An Intensional Formalization of Generic Statements.” Logica Universalis, Springer Nature, 2023. DOI
“How Combinatory Logic Can Limit Computing Complexity.” EPJ Web of Conferences 244, 2020. DOI
Pandita, Bucuvalas, Bergier, Chakarov, and Richards. “Towards J.A.R.V.I.S. for Software Engineering.” AAAI Workshops, 2018.
“Formal Logic as a Liberal Art.” In Logic in All Its Dimensions: Festschrift for Jean-Yves Béziau’s 60th Birthday, Vol. 1, Birkhäuser. In press.

Selected conferences and talks

“Verisimilitudo blandiens: Flattering Plausibility and the Epistemic Conditions of Peace.” AI4Peace 2026 International Symposium, Pontifical Lateran University, Rome, September 2026.
“What Does It Mean for AI to Understand Code?” Logic in Question (LIQ 2026), Sorbonne, Paris, May 2026.
“AI as Mere Mimicry: Why Artificial Intelligence Lacks Ethical Agency?” Computer Ethics: Philosophical Enquiries (CEPE 2025), Rome, September 2025.
1st World Congress on Logic, Chance and Money, University of Nevada, Las Vegas, November 2024.
“InterPARES Trust AI project: Can we leverage AI for RiC-O?” International study day on Records in Contexts, Archives nationales de France, Paris, November 2023.
“Can we combine logical analysis and machine learning? A NLP use case.” Datacraft Workshops, Paris, July 2022.

Editorial

Co-editor, special issue The AI Revolution: Perspectives from Philosophy of Management, Philosophy of Management (Springer), 2026.

All publications and talks in the CV →

Work With Me

Hiring or Consulting

For teams hiring

Research and applied-science roles

I am open to senior research-scientist and applied-scientist positions in AI for code, knowledge representation, and formal methods.

AI for code: program understanding and analysis
Knowledge representation and formal ontologies
Neurosymbolic systems combining LLMs with reasoning
Formal verification and theorem proving

Remote across US and EU time zones; dual US and French citizen.

Discuss a role

For projects

Consulting through Analogia Entis

Analogia Entis is my independent practice, based in France. An engagement can be a single review, a scoped piece of work, or an ongoing advisory role.

Design or audit of a formal ontology or knowledge graph
Architecture review of hybrid symbolic and LLM systems
Formalization and verification of algorithms and proofs in Lean 4
Expert assessment of the reasoning capabilities and limits of an AI system
Research collaboration and technical writing

This practice is separate from my work at Phase Change Software and does not take on legacy-code modernization.

Discuss a project

About

Logician and Computer Scientist

I hold a PhD in mathematics and formal logic from Université Paris-Sorbonne, with a dissertation on the logic of movement supervised by Jean-Pierre Desclés. I have been Research Scientist at Phase Change Software since 2017. At Regis University in Denver I was a tenured associate professor, department chair, and director of the BS in Computer Science until 2026, and I am now Visiting Professor of Computer Science there for 2026 to 2027. My courses covered computation theory (automata, recursive functions, λ-calculus, complexity), principles of programming languages, and AI ethics.

My research is in formal logic (intensional logic, combinatory logic, the λ-calculus) and in what a reasoning system can legitimately claim to know, including the rationality of large language models. Earlier, I worked as a management and strategy consultant at BearingPoint in Paris, and as a research intern in computational linguistics with Thales Group and the Université du Québec.

Full curriculum vitae →

Toolbox

TypeDB / TypeQL OWL 2 · RDF · SKOS Prolog Lean 4 · Mathlib Z3 Haskell Clojure Python Claude Code · MCP