Software engineer at Logitech working on playmaster.gg
Working on various machine learning projects in collaboration with large industrial companies involving time series data, either related to direct material procurement, industrial optimization or anomaly detection. The role notably implies the following responsibilities:
Teaching assistant to the Java programming language. Students deepen their knowledge of the Java language through a large project. They learn to use the different types of collections, design patterns and are also introduced to other concepts such as generics, anonymous classes and functions, immutability, ...
I was also in charge of helping online learners on the corresponding courses provided by the school on the Coursera platform:
Worked on distributed learning for automated machine learning pipelines at Oracle Labs. The goal was to reduce the runtime induced when training multiple machine learning models configured with different hyperparameters. The key tasks were:
Developed a prototype mobile application on Android devices to improve data collection for medical teams on the ground. The application was based on OpenMRS, an enterprise electronic medical record system. The project has been tried for a mission about children’s malnutrition in Tchad. The end goal was to help the medical staff with data collection so that they could be more responsive and apply appropriate treatments based on the patient’s medical history
Master thesis supervised by Martin Jaggi, professor in the Machine Learning and Optimization Laboratory at EPFL and Thomas Oriol, director at Datapred. My work introduces how machine learning models can be aggregated over time series data and shows how we can improve responsiveness of such aggregation algorithms when regime changes occur (grade 5.25/6).
Semester project in the Distributed Computing Laboratory at EPFL where we tested the robustness of different aggregation strategies to various attacks (for example with adversarial noisy data or gradients) by experimenting different gradient descent update rules. In particular, we test how the method developed in the lab (Krum) performs (grade 6/6).
Contributed to the proton pack library, an open source project enhancing the new Java 8 stream package.
Bachelor semester project supervised by Martin Odersky, professor and director of the Programming Methods Laboratory at EPFL and Manohar Jonnalagedda. The idea was to demonstrate that we can implement efficiently interleaved parsers in a high-level language for parsing network protocols. The slides of the presentation are available here.
Development of a video games quiz application on Android as a personal project (100 000+ downloads and 4000+ ratings).