About me

I am a Data Scientist at BCG GAMMA who graduated from the Massachusetts Institute of Technology with a Master of Business Analytics, under the supervision of Prof. Dimitris Bertsimas, with whom I did research on Machine Learning and Optimization applied to Healthcare. Before going to MIT, I obtained a Bachelor of Science in Engineering and a Master of Science in Applied Mathematics from École Centrale Paris (now CentraleSupélec, part of Paris-Saclay University). I have also interned at Wavestone’s Machine Learning & Data Lab, as well as at BCG GAMMA, both as a Data Scientist. You can find my resume here. During the past few years, I have taken part in Machine Learning and Optimization projects spanning a wide range of industries, among which Finance (Morningstar & Wavestone), Retail (BCG GAMMA), Airlines (BCG GAMMA), Transportation (Wavestone), Healthcare (MIT) and Entertainment (Comcast).

Research

My research focuses primarily on applying Machine Learning and Optimization to Healthcare problems, and the two main projects on which I have worked on were:

• A Trauma Outcome Predictor (TOP) interpretable model for patients admitted in the ICU after a trauma (blunt or penetration), where the goal was to predict the probability of mortality as well as composite morbidity (any kind of post-surgery complication) and individual morbidities. This project was done in collaboration with Medical Doctors from Massachusetts General Hospital.
• The COVID-19 pandemic, where I was the main developer of the DELPHI epidemiological model that was part, almost since its creation, of the ensemble prediction performed by the CDC. You can find articles published in the New York Times featuring my model with Michael L. Li here and here, and another one in FiveThirtyEight here. Our work has also been awarded by C3.ai’s Digital Transformation Institute ($5.4 Millions split between 26 projects) for data-driven COVID-19 research (see here).

See Research & Publications for more details.

Projects

I have worked on many Machine Learning oriented projects during my year at MIT, among which the following:

• Machines as Money Managers with Morningstar: part of the Analytics Lab class (Fall 2019), with three other students, we explored the use of Reinforcement Learning (Proximal Policy Optimization) to imitate portfolio managers in their decisions of buying/holding/selling, as opposed to a more conventional classification approach. While the code and report are the property of Morningstar, we managed to obtain an accuracy of 0.79 (vs. 0.89 with XGBoost).
• Stable Support Vector Machines and Logistic Regression: part of the Statistical Learning Theory and Applications class (Fall 2019), you can find more details about this project here.
• EVaRegression: A novel approach to Stable Regression: part of the Machine Learning under a Modern Optimization Lens class (Fall 2019), you can find more details about this project here.
• An Anchoring Approach to Medical Information Extraction using Clinical BERT Embeddings: part of the Machine Learning for Healthcare class (Spring 2020), you can find more details about this project here.
• Optimizing Content Likely Optimization with Comcast: part of our Analytics Capstone (Spring & Summer 2020), we implemented a recommender system tool using collaborative and content-based filtering methods on a dataset of 40K+ customers’ viewing profiles and 300K+ shows in order to inform marketing efforts and improve customer viewing experience. Our proof of concept showed an uplift in revenues of $5M+ as well as an improvement of 10% in accuracy to target the right customers compared to the methodology currently used.