Julien Ferry
EN FR
Get in touch

Postdoctoral researcher · Montréal

Trustworthy machine learning, grounded in optimization.

I study how machine learning models can be made more transparent and accountable—and what they can reveal about the data used to train them. My work connects combinatorial optimization with robustness, privacy, fairness, and explainability.

Explore my research Detailed CV Short CV

Currently at Polytechnique Montréal(opens in a new tab), in the Department of Mathematics and Industrial Engineering, and the SCALE-AI Chair in Data-Driven Supply Chains with Prof. Thibaut Vidal.

NeurIPS 2025Spotlight paper

ICML 2024Oral presentation

Québec ScienceTop 10 discoveries of 2024

01 / ABOUT

Short bio.

I am a postdoctoral researcher in the Department of Mathematics and Industrial Engineering at Polytechnique Montréal, in the SCALE-AI Chair in Data-Driven Supply Chains. I work with Prof. Thibaut Vidal on trustworthy machine learning and optimization.

I earned my PhD in 2023 at LAAS-CNRS / Université Toulouse III Paul Sabatier, where I studied the interplays between interpretability, fairness, and privacy in machine learning using combinatorial optimization and operations research.

My research focuses on trustworthy machine learning, with a focus on certifiability, interpretability, explainability, fairness, robustness, privacy, and the relationships among these goals. My works draw primarily on combinatorial optimization, particularly mathematical programming, to develop methods with rigorous guarantees of optimality or constraint satisfaction.

Google Scholar GitHub Detailed CV

02 / RESEARCH

Questions that drive my work.

Combinatorial optimization provides powerful and flexible tools for jointly modelling and studying many desiderata related to trustworthy ML. This allows thorough characterizations, with rigorous guarantees on optimality and constraint satisfaction.

01

Privacy & model security

What can a trained model—or an explanation of one of its decisions—disclose about its training data? I develop reconstruction attacks and study defenses for structured models.

02

Counterfactual explanations

How can we find meaningful changes that alter a model’s prediction? I study exact and efficient methods for counterfactual explanations in tree ensembles.

03

Fair & interpretable learning

How do accuracy, fairness, transparency, and sparsity interact? I design fair and interpretable learning methods and provably characterize the space of good models for several hypothesis classes.

04

Hybrid interpretable models

How to jointly leverage the predictive power of a black-box and the transparent nature of a simple model? My research studies optimization-grounded methods to learn such models and characterize the novel chanllenges they carry.

05

Faithful compression

How to compress an existing model without altering its prediction function? My work uses tailored row- and column-generation frameworks to compress existing tree ensembles while certifiably preserving their decision boundaries.

06

Robustness

How to make sure that a trained model will perform well outside its training data distribution? I use distributionally robust optimization tools grounded in combinatorial optimization to derive rigorous worst-case guarantees.

Improving one trustworthy ML desiderata is likely to affect others: only a joint handling of all desirable properties can make ML really trustworthy.

03 / PUBLICATIONS

Selected publications.

View Google Scholar
Browse the complete bibliography

The selected works above also appear below for completeness.

Journal articles

  1. Taming the Triangle: On the Interplays between Fairness, Interpretability and Privacy in Machine Learning ↗. Julien Ferry, Ulrich Aïvodji, Sébastien Gambs, Marie-José Huguet, Mohamed Siala. Computational Intelligence, 2025.
  2. Boosting Revisited: Benchmarking and Advancing LP-Based Ensemble Methods ↗. Fabian Akkerman, Julien Ferry, Christian Artigues, Emmanuel Hébrard, Thibaut Vidal. Transactions on Machine Learning Research, 2025.
  3. Learning Hybrid Interpretable Models: Theory, Taxonomy, and Methods ↗. Julien Ferry, Gabriel Laberge, Ulrich Aïvodji. Transactions on Machine Learning Research, 2024.
  4. Improving Fairness Generalization Through a Sample-Robust Optimization Method ↗. Julien Ferry, Ulrich Aïvodji, Sébastien Gambs, Marie-José Huguet, Mohamed Siala. Machine Learning, 2023.

International conferences

  1. Counterfactual Maps: What They Are and How to Find Them ↗. Awa Khouna, Julien Ferry, Thibaut Vidal. NeurIPS, 2026.
  2. Manipulation-Proof Oblivious Audits against Deceptive Model Providers ↗. Augustin Godinot, Sofiane Azogagh, Julien Ferry, Sébastien Gambs. AIES, 2026.
  3. When Interpretability Is Unequally Distributed: Fairness in Hybrid Interpretable Models ↗. Ziba Jabbar Zare, Ulrich Aïvodji, Julien Ferry, Thibaut Vidal. AIES, 2026.
  4. Training Set Reconstruction from Differentially Private Forests: How Effective is DP? ↗. Alice Gorgé, Julien Ferry, Sébastien Gambs, Thibaut Vidal. SaTML, 2026.
  5. Fairness and Sparsity within Rashomon sets: Enumeration-Free Exploration and Characterization ↗. Lucas Langlade, Julien Ferry, Gabriel Laberge, Thibaut Vidal. AIES, 2025.
  6. From Counterfactuals to Trees: Competitive Analysis of Model Extraction Attacks ↗. Awa Khouna, Julien Ferry, Thibaut Vidal. NeurIPS, 2025.
  7. Trained Random Forests Completely Reveal your Dataset ↗. Julien Ferry, Ricardo Fukasawa, Timothée Pascal, Thibaut Vidal. ICML, 2024.
  8. Probabilistic Dataset Reconstruction from Interpretable Models ↗. Julien Ferry, Ulrich Aïvodji, Sébastien Gambs, Marie-José Huguet, Mohamed Siala. SaTML, 2024.
  9. Exploiting Fairness to Enhance Sensitive Attributes Reconstruction ↗. Julien Ferry, Ulrich Aïvodji, Sébastien Gambs, Marie-José Huguet, Mohamed Siala. SaTML, 2023.
  10. Improving Fairness Generalization Through a Sample-Robust Optimization Method. Julien Ferry, Ulrich Aïvodji, Sébastien Gambs, Marie-José Huguet, Mohamed Siala. AAAI (Journal Track), 2023.
  11. Leveraging Integer Linear Programming to Learn Optimal Fair Rule Lists ↗. Julien Ferry, Ulrich Aïvodji, Sébastien Gambs, Marie-José Huguet, Mohamed Siala. CPAIOR, 2022.
  12. Learning Optimal Fair Scoring Systems for Multi-Class Classification ↗. Julien Rouzot, Julien Ferry, Marie-José Huguet. ICTAI, 2022.
  13. FairCORELS, an Open-Source Library for Learning Fair Rule Lists ↗. Ulrich Aïvodji, Julien Ferry, Sébastien Gambs, Marie-José Huguet, Mohamed Siala. CIKM, 2021.

Preprints

  1. Optimal Counterfactual Search in Tree Ensembles: A Study Across Modeling and Solution Paradigms ↗. Awa Khouna, Youssouf Emine, Julien Ferry, Thibaut Vidal. arXiv preprint arXiv:2605.06561 (under review), 2026.
  2. Pace: Prune-And-Compress Ensemble Models ↗. Fabian Akkerman, Julien Ferry, Théo Guyard, Thibaut Vidal. arXiv preprint arXiv:2605.06278 (under review), 2026.
  3. Trustworthy Machine Learning through the Lens of Combinatorial Optimization: Survey and Research Perspectives ↗. Thibaut Vidal, Julien Ferry. arXiv preprint arXiv:2607.07762 (under review), 2026.
  4. Smooth Sensitivity for Learning Differentially-Private yet Accurate Rule Lists ↗. Timothée Ly, Julien Ferry, Ulrich Aïvodji, Sébastien Gambs, Marie-José Huguet. arXiv preprint arXiv:2403.13848, 2024.
  5. Learning fair rule lists ↗. Ulrich Aïvodji, Julien Ferry, Sébastien Gambs, Marie-José Huguet, Mohamed Siala. arXiv preprint arXiv:1909.03977, 2019.

National conferences · with proceedings

  1. Exploiter l'équité d'un modèle d'apprentissage pour reconstruire les attributs sensibles de son ensemble d'entraînement. Julien Ferry, Ulrich Aïvodji, Sébastien Gambs, Marie-José Huguet, Mohamed Siala. Rencontres des Jeunes Chercheurs en Intelligence Artificielle, 2023.
  2. Améliorer la généralisation de l'équité en apprentissage grâce à l'Optimisation Distributionnellement Robuste. Julien Ferry, Ulrich Aïvodji, Sébastien Gambs, Marie-José Huguet, Mohamed Siala. Rencontres des Jeunes Chercheurs en Intelligence Artificielle, 2021.

National conferences · without proceedings

  1. Utiliser l'équité d'un modèle d'apprentissage pour reconstruire les attributs sensibles de son ensemble d'entraînement. Julien Ferry, Ulrich Aïvodji, Sébastien Gambs, Marie-José Huguet, Mohamed Siala. 24ème conférence ROADEF de la Société Française de Recherche Opérationnelle et Aide à la Décision, 2023.
  2. Concilier l'équité statistique et la précision en apprentissage machine interprétable grâce à la PLNE. Julien Ferry, Ulrich Aïvodji, Sébastien Gambs, Marie-José Huguet, Mohamed Siala. 23ème conférence ROADEF de la Société Française de Recherche Opérationnelle et Aide à la Décision, 2022.
  3. Optimisation Distributionnellement Robuste pour améliorer la généralisation de l'équité en apprentissage. Julien Ferry, Ulrich Aïvodji, Sébastien Gambs, Marie-José Huguet, Mohamed Siala. 22ème conférence ROADEF de la Société Française de Recherche Opérationnelle et Aide à la Décision, 2021.

PhD thesis

  1. Addressing Interpretability, Fairness and Privacy in Machine Learning Through Combinatorial Optimization Methods ↗. Julien Ferry. UPS Toulouse, 2023.

04 / OPEN SOURCE

Research you can use.

All the code developed for my research is made publicly available. A few examples:

COUNTERFACTUAL EXPLANATIONS

OCEAN

A library for counterfactual explanations for tree ensembles, to which I contributed refactoring and new methods.

github.com/vidalt/OCEAN PRIVACY AUDITING

DRAFT

A dataset reconstruction attack targeting the training data of random forests.

github.com/vidalt/DRAFT FAIR LEARNING

FairCORELS

Learn compact, interpretable rule lists while accounting for statistical fairness constraints.

github.com/ferryjul/fairCORELS INTERPRETABILITY

HybridCORELS

Train hybrid models that combine an interpretable component with a black-box model.

github.com/ferryjul/HybridCORELS RASHOMON SETS

Rashomon-Explorer

Precisely characterize, for several hypothesis classes, the sets of trade-offs achievable between accuracy, fairness and interpretability.

github.com/vidalt/Rashomon-Explorer PRIVACY PROTECTION

DRAFT-DP

Investigating the effectiveness of a Differential Privacy mechanism to prevent training set reconstruction attacks against random forests.

github.com/vidalt/DRAFT-DP EXPLANATIONS AND PRIVACY

Tree-Extractor

Leveraging counterfactual explanations of tree ensembles to provably recontruct the underlying model from a black-box API access.

github.com/vidalt/Tree-Extractor

More projects on my GitHub ↗

05 / EDUCATION & TEACHING

Beyond the papers.

Positions & education

  1. 2024 — present
    Postdoctoral researcher

    Polytechnique Montréal · SCALE-AI Chair in Data-Driven Supply Chains
    Supervisor: Thibaut Vidal

  2. 2020 — 2023
    PhD in Computer Science & Telecommunications

    LAAS-CNRS · Université Toulouse III Paul Sabatier
    Supervisors: Marie-José Huguet, Sébastien Gambs and Mohamed Siala

  3. Spring-Summer 2020
    Machine Learning Intern

    Thales Alenia Space · Bias mitigation in CNNs

  4. 2015 — 2020
    Engineer’s degree

    INSA Toulouse · Distributed Systems and Big Data · Ranked first in graduating class

  5. Fall 2017
    Study semester abroad

    The University of Texas at Austin · University Honors Fall Semester 2017

Teaching & mentoring

I have taught algorithms, complexity, data structures, graph theory, programming, supervised and unsupervised machine learning, and related topics at INSA Toulouse (total of 196 hours). I also gave invited lectures on AI privacy and security at UQAM (3 hours) and INSA Toulouse (1 hour). Full list of my teaching activities can be found on my CV.

I have co-supervised 8 master students, and currently co-supervise 2 doctoral students on interpretable models, counterfactual explanations, and privacy attacks.

Academic service

I have reviewed for journals including JMLR, TMLR, TKDE and Neural Networks, as well as conferences including ICML, NeurIPS (awarded Top Reviewer in 2026), AAAI, IJCAI, FAccT, SaTML, ECAI, ECML-PKDD, CP, and CPAIOR.

06 / CONTACT

Let's talk.

For collaborations, speaking invitations, or questions about my work, reach me by email.

julienferry12[at]gmail.com

Find me elsewhere

Google Scholar ↗GitHub ↗LinkedIn ↗