Beyond my research activities, I have also worked on a number of non-academic projects, including many group projects, some of which were carried out in collaboration with companies.
Forecasting CO₂ Emissions in Real Estate
I joined this project alongside my Machine Learning course in my first year of master’s, as a concrete and applied setting in which to put the concepts learned in class into practice. Over three months, in a four-person team and with weekly meetings with BNP Paribas data scientists, we studied time series to forecast future CO₂ emissions by real estate property category. Tasks included data preprocessing, feature engineering, and training models such as ARIMA, LSTM and Prophet.
Repository Project report (FR)
Optimization of an electromagnetic anechoic chamber
Given a wave-propagation PDE model in a chamber, the goal was to minimize the energy of an electromagnetic wave by determining the optimal distribution of absorbing material. The work combined a theoretical study of the model (well-posedness of the PDE system) with the implementation of numerical algorithms — finite elements for solving the PDEs, and gradient descent in Hilbert spaces for the optimization.
Repository Project report (FR)
PDE / Modeling project
Carried out during the PDE course in my first year at CentraleSupélec, where students were free to choose a problem of their own and model it with PDEs. I chose to model the growth of a spy network using a system of partial differential equations, established its well-posedness, and analyzed the influence of various parameters through numerical (finite elements) solutions.
Viral Marketing Campaign
A one-month project with the data and AI consultancy Artefact, in a four-person team, analyzing the viral dynamics of an online climate-awareness campaign created for Greenpeace. The campaign featured a fictional real-estate startup (“Orizon”) promoting seaside homes projected for the year 2100, using IPCC and NASA climate data. We studied how information spread across social networks, identified key influencers, modeled diffusion patterns, and proposed data-driven strategies to increase the campaign’s reach.
Coding Weeks
A two-week coding project at CentraleSupélec, in a five-person team. We chose to implement various user-customizable versions of Conway’s Game of Life, and then a small Python video game whose dynamics relied on it. You can play the game by running main.py in the survival-the-game-of-life folder.
Detection of anomalies in an IoT network
As a conclusion to the Information Theory / IoT course in my first year at CentraleSupélec, in a five-person team, we analyzed a log dataset following a DDoS attack to identify its characteristics. Tasks included developing machine learning tools for the detection of streaming DDoS attacks.