CV

Education

PhD in Computer Engineering, Polytechnique Montréal + MILA, 2024
M.Sc. in Computer Engineering, Polytechnique Montréal, 2024
B.Sc. in Mathematics, Université de Montréal, 2020

Work experience

March 2024 - Present: Visiting Researcher
- ServiceNow Research
- Research Director: Alexandre Lacoste
- This role marked a key moment in my research development, focusing on large-scale AI systems and LLM-based web agents. My first-author work on developing WorkArena++ has sharpened my project management skills and collaborative efforts within a large research team. A significant challenge was designing tasks representative of knowledge workers on web platforms, requiring an understanding of human cognitive processes translated into computational tasks.
Sept 2022 - Jan 2023: Master’s Research
- Polytechnique Montréal
- Supervisor: Quentin Cappart
- My first real immersion into academic research. I applied project management skills to structure the work into distinct phases—literature review, code development, theoretical framework, and results analysis. This experience confirmed my aspiration to pursue doctoral studies.
Sept 2023 - Dec 2023: Consultant
- Comité de consultation en gestion Polytechnique
- Supervisor: Mehdi Louzi
- Led technical aspects of optimizing truck routes for Intelcom, reducing required trucks by 50-75%. Managed incomplete data, developed a route optimization program, and created a visualization tool for interpreting results, improving my project management and technical communication skills.
Jan 2023 - Apr 2023: Teaching Assistant
- Polytechnique Montréal
- Supervisor: Quentin Cappart
- Teaching assistant for INF 6102, a graduate course on metaheuristics. This experience helped me develop pedagogical skills, design assignments at an advanced level, and deepened my understanding of the subject through tutorials and assignment corrections.
Sept 2022 - Aug 2023: Lead Developer – SeaPearl.jl
- Supervisor: Quentin Cappart
- Significantly improved the quality of the SeaPearl.jl project, a reinforcement learning-based constraint programming solver. Increased test coverage by 15%, created comprehensive documentation, and mentored two interns. Under my leadership, the project gained visibility with 90 GitHub stars and competed in its first solver competition.
Jan 2021 - Jun 2022: Team Lead, Data Science
- Newtrax Technologies
- Supervisor: Louis-Pierre Campeau
- Led the data science product lifecycle from concept to production. Set up the company’s first system-level reliability tests, achieving over 99.99% prediction accuracy. Managed data annotation strategy, growing the team’s product line from 2 to 10.
Sept 2019 - Jan 2021: Data Scientist
- Newtrax Technologies
- Supervisor: Louis-Pierre Campeau
- Developed the first activity detection products for the company, achieving over 99% prediction accuracy. Managed end-to-end product development, including DevOps aspects. Established coding standards and test practices for the growing team.
May 2019 - Aug 2019: Machine Learning Intern
- Newtrax Technologies
- Supervisor: Louis-Pierre Campeau
- Worked on anomaly detection in the mining sector, implementing and evaluating four deep learning models on real data. Presented results to Sandvik’s team in Finland after their acquisition of Newtrax.
May 2018 - Apr 2019: Consultant in Modeling and Simulation
- DIREXYON Technologies
- Supervisor: Kevin Lacombe
- Developed a strategic maintenance model for urban infrastructure, reducing costs by 30% while maintaining asset availability.

Publications

The BrowserGym Ecosystem for Web Agent Research

Chezelles, De, Thibault Le Sellier, Maxime Gasse, Alexandre Lacoste, Alexandre Drouin, Massimo Caccia, Léo Boisvert et al. "The BrowserGym Ecosystem for Web Agent Research." arXiv preprint arXiv:2412.05467 (2024).

Learning and fine-tuning a generic value-selection heuristic inside a constraint programming solver

Marty, T., Boisvert, L., François, T. et al. Learning and fine-tuning a generic value-selection heuristic inside a constraint programming solver. Constraints (2024).

WorkArena++: Towards Compositional Planning and Reasoning-based Common Knowledge Work Tasks

Léo Boisvert, Megh Thakkar, Maxime Gasse, Massimo Caccia, Thibault Le Sellier De Chezelles, Quentin Cappart, Nicolas Chapados, Alexandre Lacoste et Alexandre Drouin. "WorkArena++: Towards Compositional Planning and Reasoning-based Common Knowledge Work Tasks." Advances in Neural Information Processing Systems, 38, 2024.

AgentMerge: Enhancing Generalization in Fine-Tuned LLM Agents

Megh Thakkar, Léo Boisvert, Thibault Le Sellier De Chezelles, Alexandre Piché, Maxime Gasse, Massimo Caccia, Alexandre Lacoste. "AgentMerge: Enhancing Generalization in Fine-Tuned LLM Agents." NeurIPS 2024 Workshop on Adaptive Foundation Models, 2024.

Fine-Tuning Web Agents: It Works, But It’s Trickier Than You Think

Massimo Caccia, Megh Thakkar, Léo Boisvert, Thibault Le Sellier de Chezelles, Alexandre Piché, Nicolas Chapados, Alexandre Drouin, Maxime Gasse, Alexandre Lacoste. "Fine-Tuning Web Agents: It Works, But Its Trickier Than You Think." NeurIPS 2024 Workshop on Open-World Agents, 2024.

Towards a Generic Representation of Combinatorial Problems for Learning-Based Approaches

Léo Boisvert, Hélène Verhaeghe, Quentin Cappart. “Towards a Generic Representation of Combinatorial Problems for Learning-Based Approaches.” CPAIOR 2024.

WorkArena: How Capable are Web Agents at Solving Common Knowledge Work Tasks?

Alexandre Drouin, Maxime Gasse, Massimo Caccia, Issam H. Laradji, Manuel Del Verme, Tom Marty, Léo Boisvert, Megh Thakkar, Quentin Cappart, David Vazquez, Nicolas Chapados, Alexandre Lacoste, "WorkArena: How Capable are Web Agents at Solving Common Knowledge Work Tasks?" ICLR 2024 Workshop on Large Language Model (LLM) Agents.