MAPPO (May 2025, Python):
Two Multi-Agent Reinforcement Learning algorithms implemented in PyTorch and applied on the KAZ PettingZoo environment (IPPO & MAPPO).
[repo]
PettingZoo Dilemmas (March 2025, Python):
Implementation of 4 classical Game Theory matrix games using the PettingZoo environment for multi-agent reinforcement learning. Stag Hunt, Subsidy Game, Matching Pennies and Prisoner’s Dilemma.
[repo]
fastlob (January 2025, Python):
Fast & minimalist limit-order-book (LOB) implementation in pure Python. This package is meant to be safe, clean, and easy to reuse and extend for other developers. My goal with this project is that it becomes the reference package for anyone who needs a reliable limit-order-book.
[repo]
Autograd.hs (January 2025, Haskell):
Automatic differentiation engine in Haskell, inspired by Andrej Karpathy's Micrograd. The engine works only for scalar-valued functions, but I plan on integrating Jacobian computations as well.
[repo]
Binance Order-Book (November 2024, Python):
Simple project allowing one to run a Binance order-book locally in the terminal.
[repo]
Llama Aurelius (September 2024, Python, Pytorch, Transformers):
Fine-tuning Llama on Marcus Aurelius' writings using QLoRA. The resulting model is available on my HuggingFace profile.
[repo] [model]
MLE Linear Regression (July 2024, Python):
Brief Jupyter notebook for how we can derive the Ordinary Least Squares formula from a normal (gaussian) distribution Maximum Likelihood Estimation.
[repo]
NN.c (July 2024, C):
Feedforward neural network (MLP) & other ML models implemented from scratch in pure C.
[repo]
Summer Research Internship (May/June/July 2024, Python, PyTorch, Transformers):
My second research internship at the LaBRI (Bordeaux Computer Science Laboratory). This time, I focused on a project at the crossroads of Deep Learning and Software Engineering. My work consisted in finding ways to train and use Large Language Models more effectively for Software Engineering. I learned how to train Transformers from scratch, as well as how to fine-tune or internally modify pre-trained models and conduct experiment on them. If you understand French, I gave a quick talk about my work at the end of my internship, and this presentation was filmed.
[presentation] [repo]
FuncSkeleton (June 2024, Python):
Python package to generate the CFG "skeleton" of a function, for Machine Learning purposes. It is a tool built during my internship that I used to create a dataset from CodeSearchNet.
[repo]
MMA Predictions (April 2024, Python, Pandas, Sklearn, PyTorch):
Using various machine learning algorithms on UFC fights data to try to predict outcomes such as the winner, the result or the finish round.
[repo] [paper]
Threaded Thrift Store (April 2024, Java):
Concurrent thrift store simulation in Java.
[repo]
Three-Address Code Compiler (February 2024, Java):
Compiler taking an hypothetical (simplified) programming language source code file and converting it to three-address code for a given interpreter.
[repo]
Mail.tm (January 2024, Python):
Simple wrapper around the Mail.tm API in Python to interact easily with it.
[repo]
Feistel Network Framework (September 2023, Python):
Feistel network framework implementation where users can provide their own round and key derivation functions.
[repo]
Market Exchange (August 2023, Go):
Market exchange simulation project where users can place (limit) and execute (market) orders via an API.
[repo]
CRT (July 2023, OCaml):
Simple Chinese Remainder Theorem implementation for solving systems of modular congruences.
[repo]
LZSS (May 2023, C):
Lempel–Ziv–Storer–Szymanski algorithm implementation in C for file compression & decompression.
[repo]
Busy Beaver (April 2023, Go):
Simple 2-state Turing machine implementation in Go to experiment on the busy beaver problem. It is as an appendix for a small article (in french) that I wrote as part of an assignment for my university.
[repo]
Advent of OCaml (March 2023, OCaml):
My attempt to Advent of Code 2023 using only OCaml.
[repo]
Huffman Coding Calculator (March 2023, C++):
Huffman coding gain calculator for text files, it encodes a given file and computes the difference of size.
[repo]