I am a PhD candidate in Computer Science in the Department of Computer and Information Sciences at the University of Delaware working under the supervision of Professor Beheshti within the healthy lAife lab.
I received my Bachelor of Engineering Science and my Master of Science in Engineering (Computational Finance track) from EFREI Paris in 2018 and 2020. When I'm not working you can find me playing tennis, taking photos, watching the Montreal Canadiens, or cheering for the French National Football (soccer) team.
My research iterests lie at the intersection of Machine Learning and Healthcare. Specifically, I am interested in developing robust Deep Learning techniques for Electronic Health Records to tackle problems related to Fairness, Missing Labels and Few-Shot Learning, Multi-Task Learning, or more generally, Cardiovascular Disease.
Our preprint, "Aligning (Medical) LLMs for (Counterfactual) Fairness" is available on arxiv. [Paper]
Our preprint, "Fairness-Optimized Synthetic EHR Generation for Arbitrary Downstream Predictive Tasks" is available on arxiv. [Paper]
Our preprint, "Bias patterns in the application of LLMs for clinical decision support: A comprehensive study" is available on arxiv. [Paper]
Our paper, "Graph Transformers on EHRs: Better Representation Improves Downstream Performance" has been accepted in ICLR 2024. [Paper]
I will be joining Pinterest (PinLabs) as an ML intern within the Inclusive AI team for the Summer 2023.
Our paper, "Improving Fairness in AI Models on Electronic Health Records: The Case for Federated Learning Methods" has been accepted in FAccT 2023. [Paper]
Our paper, "An Extensive Data Processing Pipeline for MIMIC-IV" has been accepted in ML4H 2022. [Paper]
Our paper, "Few-Shot Learning with Semi-Supervised Transformers for Electronic Health Records" has been accepted in MLHC 2022. [Paper]
Our paper, "Flexible-window Predictions on Electronic Health Records" has been accepted in IAAI 2022. [Paper]
Our paper, "Transformer-based Multi-target Regression on Electronic Health Records for Primordial Prevention of Cardiovascular Disease" has been accepted in BIBM 2021. [Paper]
Joined the healthy lAife lab @ University of Delaware as a PhD Student.
I built an end‑to‑end pipeline to quantify and analyze demographic biases in text‑to‑image generative models, and proposed multiple solutions to mitigate biases issues such as Textual Inversion, Prompt Engineering, and fair fine‑tuning through LoRA.
I built a testing automation tool using Python and Cucumber to help the Data Shop team running their unit tests during the development of the company's new on-demand market data platform.
I built an architecture dataflow visualization tool for their trading platform, Optiq to help engineers visualize how each applications communicate with each other and follow a specific data stream from its creation to its endpoint.
Teaching Assistant