Software
Below are several open-source projects I have architected and actively lead as part of my work, in collaboration with colleagues and community contributors.
🩺 EDS-NLP 
Modular, fast NLP framework, compatible with Pytorch and spaCy, offering tailored support for French clinical notes.
🛡️ EDS-Pseudo 
EDS-Pseudo is a hybrid model for detecting personally identifying entities in clinical reports. We also provide a pretrained, open-source model as part of this project.
📄 EDS-PDF 
EDS-PDF is a generic, pure-Python framework for text extraction from PDF documents. It provides the machinery to use rule- or machine-learning-based approaches to classify text blocs between body and meta-data.
⚙️ Confit 
Confit is a complete and easy-to-use configuration framework aimed at improving the reproducibility of experiments by relying on the Python type annotations, minimal configuration files and a robust CLI.
📦 FoldedTensor 
A PyTorch extension that provides efficient handling of tensors containing deeply nested sequences variable sizes.
Here are some personal side projects I have developed in my spare time or as part of my PhD thesis.
⚛️ Pret 
Library for building full-stack reactive user interfaces in Python, using React as a rendering engine. This project is under active development and likely not ready yet for production use.
🏷️ Metanno 
Annotator building tool for JupyterLab, enabling customizable and interactive text annotation directly in Python.
📊 Rich Logger 
Table logger using Rich, designed for logging training metrics in PyTorch Lightning / Accelerate with customizable formatting.
🔄 PyGetSource 
A decompiler for Python 3, converting compiled bytecode back into readable Python source code.
🏗️ NLStruct 
(Archived) A library for building structured prediction models in NLP, developed as part of my PhD thesis. It focuses on named entity recognition and entity attribute extraction. I now focus my efforts on EDS-NLP.
💻 SublimeTerm 
(Archived) Sublime Text plugin for emulating a terminal inside a tab, enabling shell access without leaving the editor.