Research Projects

Study of notions of fairness and conception of algorithms in multi-criteria collective decision-making problems. Including fair allocation of indivisible goods, with rich preference structures (Matroid Rank Preferences, etc.). This implies equilibria study (game theory), design of strategy-proof mechanisms, and approximation algorithms for NP-hard problems. The work is theoretical, with some experiments on synthetic data.

Modelling EU railway networks (Switzerland, France) and evaluating fairness using social choice metrics — p-Social Welfare, Gini and Atkinson indices — based on travel time and distance relative to ideal lines. Networks are available on the Railway Networks page.

Digital urban twin of Hamburg for sustainable urban development. Agent-Based Modeling in GAMA Platform, in collaboration with Hamburg's Ministry of Ecology and Housing. The model compares retrofitting versus demolition-reconstruction policies, revealing that demolition only breaks even ~50 years later.

Simulation and resolution of spatial conflicts between pedestrians using social choice theory. Developed a Python/Mesa platform (CNM — Conflict Negotiation Model) outperforming SFM and CBS-MAPF in conflict resolution.

Proposal and implementation of Java type representations in the Flix programming language. Work done directly in the Flix compiler (Scala 2): new syntax parsing, type reduction, and JVM-based operations.

WebAssembly serverless join vs. native performance. Built a Rust comparison platform across serverless/serverful environments. Results: WebAssembly is up to 2x slower than native, sometimes only 20% slower, with lower cold-start times as a notable advantage. Open perspectives for optimizing WebAssembly performance include Wasm modules management as decentralized programs, within a sandboxed environment.

Database architecture (MySQL, Pandas) to process French public procurement data and detect potential frauds. Developed an LLM-based assistant prototype for lawyers to pre-process and support decision-making. Work is private; a press article covers the project.

Simulation of a generic negotiation model based on Earnest's model and Putnam's two-level game theory. Results: more agents -> harder negotiations, contradicting Earnest's proposition. Fully customizable for n-choice problems; batch experiments using genetic algorithms.