I am a french PhD student in AI specialized in LLM for Code Generation. My research is about building Software Engineering Agents (SWE-Agents), i.e. LLM-based systems interacting with a codebase through a terminal interface to solve a long-term coding task, such as solving a GitHub issue, an embodied environment or even lead a research project. I believe this framework can tackle a vast quantity of tasks and is extremely promising for the future of AI.

I have a strong background in mathematics and formalization from my french engineering masters. I'm mostly a python developper and not an expert on other languages, but I compensate by heavy but carefully designed AI coding assistant usage. I have teaching experience in RL, RL on LLMs, and computer science project management.

I have also a technical and conceptual interest on other AI fields such as Supervised Learning, Reinforcement Learning, Game Theory/Game Playing AI, Operational Research, Search Problems/Decision Making, Social Deduction Games benchmarks, as well as a philosophical and conceptual interest for AGI and AI Alignment & Safety.

• Boulet, T., Hinaut, X., & Moulin-Frier, C. (2025). Software Engineering Agents for Embodied Controller Generation: A Study in Minigrid Environments. NeurIPS 2025 Workshop on Efficient Reasoning.

• Taylor-Davies, M., Hamon, G., Boulet, T., & Moulin-Frier, C. Emergent kin selection of altruistic feeding via non-episodic neuroevolution. International Conference on the Applications of Evolutionary Computation (Part of EvoStar 2025). 👑 EvoApps best paper award, EvoStar best student paper award.

• (Medium Article) Deep Reinforcement Learning for Network Design in Marine Transportation

(Details on my LinkedIn)

• PhD at FLOWERS: Using Large Language Model for generation of controllers in interaction agent-environment - current

• Research Intern at FLOWERS (INRIA's Reinforcement Learning team in Bordeaux (France)): Studying the dynamics of Ecosystems of Neural Agents - 6 months

• Maths Vision Apprentissage (MVA) master at Ecole Normale Supérieure Paris Saclay : one of the most renowned masters in Europe for machine learning - 1 year

• R&D intern at Ubisoft La Forge Montréal: Deep Learning applied to Video Games - 6 months

• Research intern at InstaDeep Paris : Deep RL for optimization in maritime Network Design - 6 months

• Student at CentraleSupélec : top 3 French Engineer School (Deep Learning lessons and many more) - 3 years

• Courses and Practical Works on Reinforcement Learning at Ecole Nationale Supérieure de Cognitique Bordeaux. Teaching materials here.
• Courses on Reinforcement Learning as part of the CentraleSupélec AI student association.
• Template : a template for starting Research Projects with good code practice and usefull tips.
• Web App : A playground for Reinforcement Learning algorithms. It is a simple app that allows you to play with different environments, algorithms, and hyperparameters.
• A Gridworld Reinforcement Learning Environment and Framework.

• Alan Code : An open-source coding agent. Usable in CLI, GUI, or as a python library to build upon.
• Reasoning Game Act framework : A framework to benchmark agents (Alan Code) or AI assistants (Claude Code, Codex) on interactive game solving tasks (ARC AGI 3, Minigrid, ...), with harness features.
• SimpleLM : My own LLM inference engine to support models, largely vibe-coded.

• fileproxy : a python package to establish file-based communication between a server and a client, as a workaround for cluster without compute node internet access. PyPI package : fileproxy
• Time measure package : a simple context manager to measure the time taken by a block of code, that only add one line of code per section to measure. PyPI package : tmeasure
• VM Placement problem : The formulation sof the VM Placement problem and several variants of it, as well as algorithms for solving them.
• Deep RL Framework : a framework made by myself for Deep Reinforcement Learning. It is based on the OpenAI Gym and pytorch.
• AI agents for Search Problems : a python package for Search Problems and Search Algorithms (DFS, A*,...). PyPI package : SearchProblemsAI.
• AI agents for Games : a python package for Perfect Information Extensive-from Games (Chess, Connect4,...) and some algorithm ((Expecti)MiniMax, MCTS, ...).
• LocalPerf : an easy-to-use package for measuring the performance of python in any machine, in terms of CPU, multiprocessing and GPU (pytorch and JAX), and also verify that the GPU is used. PyPI package : localperf.
• Telegram bot : Sold-out Tickets Detector. Repository here.
• IntegOR : a python package based on scipy.optimize that aims to traduce an Operation Research problem that can be linearized to a binary integer linear problem, to solve it. PyPI package : integor.
• Stable Baselines 3 Full Pipeline. A repository built on top of SB3 with some additional features. Repository here.
• Backtracking : A python package for backtracking algorithm, as an interface and a solving function. PyPI package : backtracking.
• Computer Vision Project : Facial Expression Recognition. Information here.

• Research project with Google DeepMind : Population-based methods for Multi-agent Reinforcement Learning, studied in the minimalistic framework of Normal Form Games.
• Machine Learning CentraleSupélec 2022 Competition : our team ranked 1st out of 76 teams. Our repository here.
• QRT 2024 Data Challenge : A competition about guessing Football matches winners. Our team of 2 ranked 2nd/24 of the Academic leaderboard, and 8th/387 of the Private leaderboard.
• School Research Project : PCA-guided K-means. A paper-reproduction benchmark on various K-means initialization method, in particular PCA-guided search.
• School Project : Improved Semantic Segmentation for Night Drive (mission for Valeo).
• School Project : Task Scheduling with A* and IDA*. Implementation of an informed search algorithm A* and IDA* for solving a scheduling problem.

You can contact me on LinkedIn or by mail.

Personal Mail : timothe.boulet0@gmail.com
INRIA mail : timothe.boulet@inria.fr