R&D Contributions:- Designed and developed a model using a multimodal graph neural network (protein and molecule) to predict bioactivity in zero-shot learning. As project lead, I led a team of 6 individuals throughout the entire process, from conception to implementation.- Conceptualized and implemented a complex architecture using a geometric graph neural network to determine the 3D pose of a molecule based on a protein. As project manager, supervised a team of 3 individuals, overseeing the project's development from start to finish.- Innovated a novel framework for predicting and generating new peptides, including the design of a new type of MORGAN for proteins and peptides. As project lead, I collaborated with a team of 2 individuals to realize this innovation.- Developed a Large Language Model to leverage scientific annotations of molecules, contributing to advancements in computational chemistry research.- Designed and implemented a framework for automating the hyper-optimization of machine learning models to predict various properties of molecules, including bioactivity and ADMET. This framework was successfully deployed in production.- Led the development of a deep learning framework using a generative model to propose new molecules based on a multi-parametric score. Designed new reinforcement and genetic algorithms to enhance this framework, which is widely utilized in production.Other Contributions:- Participated in around twenty service projects as a data science support for pharmaceutical projects.- Created educational materials and conducted seminars to promote the latest advancements in deep learning, facilitating knowledge dissemination within the team.- Mentored and supervised 6 interns and 3 junior data scientists, guiding them in their professional and technical development.

- Developed a predictive model aimed at preventing failures and downtime of BNP Paribas IT equipment.- Integrated data from various departments using Kafka and Spark for improved data processing and analysis.

- Conducted customer segmentation utilizing clustering methods, deployed in production.- Developed Key Performance Indicators (KPIs) aligned with marketing and accounting business rules, deployed in production.- Created a predictive model for customer churn rate using Random Forest algorithm.

- Designed and developed an innovative scientific program enabling material simulation using distributed and shared parallelization techniques to study highly correlated materials.- Utilized Quantum Monte Carlo methods to solve quantum equations considering the N-body problem with an electronic wave function, contributing to a deep understanding of material properties.- Analyzed and leveraged the wave function to deduce interpretable physical phenomena, enabling precise interpretation of simulation results.- Exploited results from the equation of state to identify a first-order phase transition between two phases with the same crystalline structure but different electronic configurations.- Developed statistical models in Fortran and Python to extract meaningful information with physical interpretations, facilitating the analysis and interpretation of data generated by simulations.- Disseminated discoveries and results through seminars, scientific articles, and a thesis report, thereby contributing to the advancement of knowledge in the field of materials science.