Martin Beaussart

Experience

Volv Global

Data Scientist, January 2023 - present

Develop AI models to diagnose people with rare diseases using Electronic medical records (EMRs).
Create and design data analytics to provide customers with valuable patient and disease information, such as the patient journey.

Centre hospitalier universitaire vaudois (CHUV)

Data Scientist in the Oncology Department, January 2022 - August 2022

Created a method to automatically label cells in hematoxylin and eosin (H&E) images.
Improved model accuracy with an average of 94%, compared to a maximum of 78% at the beginning of the project.
Improved project time complexity from exponential to linear, resulting in a 27 times faster process.

Portfolio

Cell type identification using Deep Learning Model trained with cell type information transferred from mIF to co-registered H&E images

This project was conducted at the Centre hospitalier universitaire vaudois (CHUV). The goal was to automatically label H&E cells by transferring the information from the mIF WSI. The transfer was achieved by doing a registration on the two images. The nearly one million cells extract from three tissues allowed me to train different deep learning models, giving me an accuracy of around 93%.

Tools used : Pandas, Pytorch, SimpleElastix, Skimage, Numpy and Openslide

Links : pdf

Automate personal finance reports

Project of a personal finance management tool that automates the process of cleaning, categorizing, and reporting bank transactions. The project takes a dataset of monthly bank transactions as input, applies data cleaning and categorization using transformer models, and generates a comprehensive report.

Tools used : Pytorch and Plotly

Link : Github

Exploring the crimes in Chicago from 2001 to 2019

Visualization project, the goal of which was to provide an overview of all types of crimes between 2001 and 2020 in the city of Chicago. This is a creative project where I developed visualizations mainly using D3js.

Tools used : Pandas, D3js, Leaflet and Jawg

Links : Website

Personalized Federated Learning - benchmarking the weight erosion aggregation scheme

Implementation of different personalized federated learning methods. The goal was to evaluate the weight erosion model with different successful aggregation schemes on a complex task, COVID-19 detection with ultrasound images.

Tools used : Pandas, Pytorch and Numpy

Link : Github

Publication

Martin Beaussart, Felix Grimberg, Mary-Anne Hartley, Martin Jaggi (2021)

WAFFLE: Weighted Averaging for Personalized Federated Learning

NeurIPS 2021 Workshop on New Frontiers in Federated Learning: Privacy, Fairness, Robustness, Personalization and Data Ownership

Links : arXiv and NeurIPS paper

Nikhil Khandelwal, Sally Higgins, Martin Beaussart, Vahid Esmaeili, Christopher M Rudolf, Jimmy Hinson (2024)

Overcoming Racial and Ethnic Biases in the Diagnosis of Patients With Alpha-1 Antitrypsin Deficiency in the United States Using a Machine-Learning Model

American Thoracic Society 2024 International Conference

Links : Poster

J.K. Stoller, N. Khandelwal, V. Esmaeili, C. Blair, M. Kirby, J. Hinson, L. Jehi, J. Shen, M. Beaussart, D.C. Falcón, L. van Wouwe, and U. Hatipoğlu (2025)

Application of an Artificial Intelligence Model to Detect Alpha-1 Antitrypsin Deficiency: Characterizing the Study Population

American Journal of Respiratory and Critical Care Medicine

Links : Paper