Postdoctoral position funded by the European network of excellence ENFIELD, of which the Institut Mines-Télécom (IMT) is a member. Worked with Sébastien Harispe (Professor in Computer Science at IMT Mines Alès) and Nicolas Soulié (Associate Professor in Economics at IMT Business School). Studied the impact of explainable AI (XAI) techniques on human-computer collaboration, including the design and implementation of a large-scale experimental protocol involving a cohort of 600 human participants.

GIRO is one of the world leaders and the European leader in software solutions for organizing public transportation (bus and rail networks). I worked as a developer in the Algorithms team, where my role was to develop and maintain combinatorial optimization algorithms in C++ for bus and train driver rosters, in particular metaheuristic algorithms relying on linear programming solvers.

Title: "Combinatorial search guided by machine learning for molecular chemistry". Computer Science specialty. Defended on December 9, 2022, the manuscript being available on HAL. Supervised by Béatrice Duval (director, Professor Emerita, LERIA), Benoit Da Mota (Associate Professor, LERIA) and Thomas Cauchy (Associate Professor, MOLTECH-Anjou), all at the University of Angers. Reviewed by Laetitia Jourdan (Professor, CRIStAL, University of Lille) and Jean-Claude Crivello (Research Scientist HDR, ICMPE, CNRS), with Gaël Varoquaux (Research Director, INRIA Saclay) as jury president.

Decisional Intelligence track (research-oriented Master's program). Includes courses in combinatorial optimization and machine learning. Graduated in August 2019 (with highest honors). Master's thesis: "Predictive models for molecular quantum chemistry computations".

Bachelor's in Science, Technology, Health, Computer Science major. Graduated in August 2017 (with honors).

  • Supervision of a group of 3rd-year engineering students working on the development of explainable predictive models for the prediction of acute myeloid leukemia mortality, in partnership with the University of Lille.
  • Proposed and supervised a 3-month research internship for a 2nd-year engineering student, consisting in training predictive models and implementing explainability techniques for my research work.
  • Co-led a scientific activity related to my research for 3rd-year engineering students during their Deep Learning course, consisting in a complete run of my experimental protocol on the web platform, followed by a presentation of the scientific stakes of the experiment and of the implementation of the models and explainability techniques.

  • 192h of teaching over three years (maximum allowed: 64h/year), mainly to BSc and MSc students.
  • Subjects: algorithmics, Culture numérique (BSc level), and design patterns (MSc level).
  • Main responsibilities: leading lab sessions, contributing to the design and grading of assessments.
  • Supervision of group research projects for final-year BSc students.

  • 132h of tutoring over two years within the Computer Science department of the University of Angers.
  • Open-access support sessions for BSc students in Computer Science (120h).
  • Occasional assistance to Culture numérique instructors for students across the Faculty of Sciences (12h).

• GDR RADIA – Reasoning, Learning and Decision in Artificial Intelligence (working group EXPLICON).
• GDR IHM – Human-Computer Interaction (working group Human-centered AI).

A total of 5 reviews for the following journals and events:

• Scientific Reports (2)
• Journal of Cheminformatics
• SoftwareX
• Workshop STRAI@ARES2025

Member of the program committee of the Workshop STRAI held at the ARES2025 conference.

Participation in the PINKCC Challenge for image segmentation of ovarian cancer with an IMT Mines Alès team, where we obtained the 3rd prize.
Annual challenge organized by the Institut du Cancer de Montpellier, dedicated to image segmentation in radiology. The 2025 edition focused on ovarian cancer. I took part in the challenge and in the final event (13 June 2025) as a member of the IMT Mines Alès team led by Gustavo Andrade Miranda. We obtained the third place among the 20 teams selected for the final.

Best applicative article prize at the conference EGC2019, for our article "Des réseaux de neurones pour prédire des distances interatomiques extraites d'une base de données ouverte de calculs en chimie quantique".
Article on the prediction of interatomic distances with neural networks, from quantum chemistry data.

2025    github.com/PAJEAN/WebXAII
Open-source web application developed in Javascript for the study of human-XAI interaction. It supports complete experimental protocols: instructions, decision tasks involving predictive models and explanations, and questionnaires. It collects the results and records the actions of the participants.
Contribution: definition and management of the project, supervision of the research engineer in charge of the development.

2021    github.com/jules-leguy/BBOMol
Open-source tool developed in Python, combining the EvoMol evolutionary algorithm with a surrogate model (Gaussian process regression) for the optimization of molecular properties whose evaluation is costly. The surrogate model guides the search by selecting promising candidates, thus limiting the number of calls to the objective function.
Contribution: I entirely developed the tool and I maintain it.

2020    github.com/jules-leguy/EvoMol
Open-source tool developed in Python for the generation of molecules by exploring the space of molecular graphs. It allows the optimization of arbitrary target molecular properties using simple and generic mutation operators, and provides a tree-based visualization of the search for interpretability purposes. It has become a reference of the literature for the optimization of molecular properties, and is still maintained, used and cited in 2026.
Contribution: I entirely developed the tool and I maintain it.

2020    sourceforge.net/projects/mixonat
Open-source tool developed in Python for the identification of chemical compounds in a mixture from NMR spectra. Developed in collaboration with researchers in plant chemistry at the University of Angers. The tool has become a reference of the literature in the field of dereplication.
Contribution: I formalized the main algorithm, improved its algorithmic complexity, and refactored the implementation according to better programming standards.