Projects
ProteinTuneRL
ProteinTuneRL is a flexible framework for protein sequence optimization using infilling language models (e.g., IgLM) and reinforcement learning. It includes built-in tools for antibody engineering, allowing users to easily optimize protein sequences for specific tasks. The framework supports various reinforcement learning algorithms and provides a user-friendly interface for training and evaluating models.
🔗 GitHub Repository
📅 November 2024 – Present
Protlib Designer
Protlib-Designer is a lightweight Python library for designing diverse protein libraries by seeding linear programming with deep mutational scanning data (or any matrix-formatted score data). Given a score-matrix (rows = mutations, columns = score sources), it Pareto-minimizes across sources while maximizing library diversity.
🔗 View on PyPI · GitHub Repository
📮 Post
📅 November 2024 – Present
Deep Symbolic Optimization
Deep Symbolic Optimization (DSO) is a deep-learning framework for discrete optimization tasks. DSO has been designed to be extensible and modular, allowing for the easy addition of new symbolic optimization tasks, specially in the space of Reinforcement Learning for Symbolic Mathematics. The package includes the core symbolic optimization algorithms, as well as support for two particular symbolic optimization tasks:
- Symbolic Regression – recovering tractable mathematical expressions from data
- Symbolic Control – discovering interpretable policies for reinforcement-learning environments
🔗 GitHub Repository
📅 January 2021 – Present
Cardiac Machine Learning
Cardiac Machine Learning is a repository that provides code for reconstructing cardiac activation maps and transmembrane potentials from ECG signals. The code is based on the work presented in the paper “Cardiac Activation Map Reconstruction from ECG Signals Using Deep Learning” (2020). The repository includes a dataset of synthetic and real ECG signals, as well as pre-trained models for both tasks.
🔗 GitHub Repository
🔗 Cardiac Challenge
📮 Medium Post
📅 July 2018 – July 2021
Cardioid
Cardioid is a comprehensive cardiac multiscale simulation suite spanning biological scales from subcellular mechanisms to organ-level and clinical phenomena. The suite provides tools for simulating cardiac electrophysiology, cardiac mechanics, and torso ECGs, and includes utilities for cardiac meshing and fiber generation. A core component of Cardioid is its use of the finite element method for spatial discretization of cardiac tissue and torso geometries, enabling high-fidelity simulations of electrophysiology, mechanics, and ECG forward models on realistic anatomical meshes.
🔗 GitHub Repository
📅 2018 – 2020
EXIFSI
EXIFSI (EXplicit coupling schemes for Incompressible Fluid-Structure Interaction) is a four-year research project funded by the French National Research Agency (ANR) through the JCJC program. EXIFSI focuses on the development and analysis of efficient explicit coupling schemes for incompressible fluid-structure interaction problems, a challenging class of multi-physics problems that arise across engineering and biomechanics. The project’s core scientific objective is to marry decoupled time-marching strategies with stability and numerical accuracy, addressing the computational stiffness introduced by fluid incompressibility and enabling more efficient simulations of systems where deformable structures interact with surrounding fluids.
🔗 Project Website
📅 August 2012 – July 2016 (4-year ANR project)
📌 Funded by ANR JCJC (French National Research Agency)