This page is under construction. You can also view the list of my projects and publications in my Curriculum Vitæ.
We have modelled a few railway networks of EU countries, including Switzerland and France available on this website.
We evaluated fairness on railway examples (i.e., made-up graphs) and real railway networks. We define fairness with social choice metrics, e.g., p-Social Welfare, Gini or Atkinson indices, based on travel times and distances compared to ideal railway lines, i.e., time and distance of travel are proportional.
Deeper analysis will be available in the report, currently in progress, which will also be submitted to a conference.
Proposed a digital urban twin of Hamburg towards decision-making help in sustainable urban development. Using Agent-Based Modeling simulation in GAMA Platform, in collaboration with the ministry of Ecology and Housing of Hamburg.
The model gives insights on custom-set policies on retrofitting versus demolition-reconstruction of buildings. For example, it could be considered to demolish and rebuild everything (let alone socio-economic issues) for this solution to be worth only 50 years later compared to retrofitting of buildings...
Simulation and Resolution of Spatial Conflicts between Pedestrians based on Social Choice Theory.
I developed a Python simulation platform to simulate pedestrians and their collisions, with Mesa. With foundations from social choice and game theory, they simulate pedestrian or robot motion.
Our proposal, CNM (Conflict Negotiation Model), works in discrete setting and showed great performance in conflict resolution compared to the Social Force Model (SFM) and Conflict-Based Search (CBS) MAPF.
Paper published at JFSMA 2025, France.
Proposal and implementation of Java type representations in the Flix programming language.
I have worked directly in Flix's compiler (in Scala 2) to implement the use of Java in Flix, based on the JVM. This implies parsing a new syntax, type reduction and JVM-based operations.
I did this research internship because I wanted to discover other research aspects of computer science, including programming languages design.
This is a research project in parallel to my M1 at ENS Rennes.
We study the WebAssembly serverless join compared to native performance. We proposed a platform to simulate such comparisons in different environments (serverless, serverful, ...) in Rust.
Insights revealed potential in using WebAssembly in such environments (up to max 2x slower than native, sometimes 20% slower only) but requiring different approaches, notably by exploiting lower cold-start times.
Paper published at SSDBM 2024.
I chose this topic because I wanted to discover another research aspect of computer science, in particular with a private company.
This is a research project I chose to participate in, in parallel to my M1 studies.
We developed a database architecture to process French public markets data, using mySQL, and Pandas in Python. The goal is to study (potential) frauds and help lawyers analyze them.
Proposed a first version of a tool assistant for lawyers using LLMs to pre-processess requests for decision-making support.
Anticor is a French association fighting corruption in France. Unfortunately, the work is private and there is only a subset of a press article talking about our work.
I developed a simulation platform of a generic negotiation model based on Earnest's model and orginally inspired from Putnam's two-level game theory. It is developed with Mesa in Python.
The project recalled the stochastic nature of MAS, in particular the model highlighted that the more people there is the harder negotiations are, contrary to Earnest's proposition.
The model can be tweaked (tweakable model is the name of the model...) and almost fully customizable for n-choice generic problems. We run extensive batch experiments based on genetic algorithms to find solutions in the parameters' space.