Master of Science (MSc) in Data Science

Hi, I'm Jérémy.

I recently graduated with a Master of Science (MSc) in Data Science from the Swiss Federal Institute of Technology in Lausanne (EPFL), with a solid foundation in engineering, mathematics, and theoretical concepts.

Get in touch

GitHub ↗LinkedIn ↗Email ↗

❯ generative ai — live demo

❯

↳

— About

BASED IN: Lausanne, Switzerland
STUDYING: Data Science, EPFL
YEARS: 5
PROJECTS: 11+

What motivates me most is research, exploring new questions, learning new approaches, and tackling challenging problems. Having gained academic experience, I am now looking to contribute to building large-scale, open, and safe AI systems.

[email protected]

— Skills

Tools I reach for.

Languages

French

Native

German

Native, level C1

English

Upper-Intermediate B1/B2

—2018 - 2019 : 2 intensive English seminars (1 week each) in Ireland at the Galway Cultural Institute (GCI)
—2018 : Passing of the language certificate at the Goethe-Institut : Level C1 (80/100)
—2016 : Integration of the European German class.

— Journey

Education & experiences.

Sep 2026 · completed

Master of Science (MSc) in Data Science

EPFL - Lausanne

Swiss professional title: Ingénieur en Data Science (ing. data sc. dipl. EPF) · Grade: 5.64/6

Feb - Aug 2026experience

Master thesis — ML R&D Intern at QUALCOMM

Prof. Temko and Prof. Bosselut

Developing an LLM-based CWE detection pipeline for RTL code using graph retrieval, fine-tuning, and ML classification. Final grade: 6/6.

Feb 2026experience

First paper published: AI meets Mathematics in Education ↗

Prof. Friedli, Prof. Bosselut, Dr. Sotnikova

CHI 2026

2025experience

Teacher assistant in Modern Natural Language Processing

Prof. Bosselut - CS-552

2024experience

Technical supervisor for the CaLlm project. Maintaining the course and LLM infrastructure for more than 700 students: ↗

Prof. Friedli and Prof. Bosselut

2024experience

Teacher assistant in Data Intensive Systems

Prof. Kashyap - CS-300

2024 · started

Master of Science (MSc) in Data Science · EPFL

Jul 2024 · completed

Passing of the Bachelor in Computer Sciences

EPFL - Lausanne

Grade: 4.98/6

2023 and 2024experience

Teacher assistant in Applied software engineering

Prof. Mathis

2023experience

Teacher assistant in Theory Of Computation

Prof. Göös - CS-251

2021 - 2022experience

Coach for first-year students

EPFL - Lausanne

Sep 2020 · started

Bachelor in Computer Sciences · EPFL

2020diploma

Passing of the Baccalauréat in science engineering. Honors: Very good

Don Bosco - Landser, FR

2017diploma

Passing of the brevet des collèges. Honors: Very good

Don Bosco - Landser, FR

1 week - 2017experience

Internship in Cryostar

Huningue, FR

2012 - 2017diploma

Don Bosco Secondary School

Don Bosco - Landser, FR

2006-2020experience

Sports

14 years of Basketball; Scuba diving (level 1); Skiing

— Projects

Things I've built.

CaLlm (Calculus LLM)

Designed, trained, and deployed a specialized calculus model to assist EPFL students by providing targeted hints that guide them toward the correct answer without revealing full solutions. We created and translated the dataset, augmented it with reasoning traces, and trained eight models across different families (up to 14B parameters) using EPFL clusters and 16 H100 GPUs. The system is now integrated into the course platform as the default first response to students, with professors able to endorse or adjust the model's output. Currently used by over 700 EPFL students, the project remains under active development and is supervised by Prof. Sacha Friedli and Prof.

2nd place at the Explainable AI in cancer medicine competition hosted at IEEE-SSCI Trondheim. Invited to show our work.

We had the chance to present our project on explainable AI for personalized cancer medicine at IEEE SSCI 2025, alongside Jérémy Barghorn and Jérémy Chaverot. This was part of a Kaggle competition organized by Prof. Arvind Mer, which we joined during the Machine Learning course at EPFL taught by Prof. Martin Jaggi and Prof. Nicolas Flammarion. We were introduced to the competition by Prof. Jasmina Bogojeska (ZHAW), who kindly supervised our work. Out of 18 teams, we were selected among the five finalists to present at the conference and ended up 2nd!

Eliciting Reasoning in LLMs using Logprob-based Rewards

Despite larger models and Chain-of-Thought prompting, LLMs still falter at consistent reasoning. Building on DeepSeek-R1 and GRPO, we train models to trust their own confidence using logprob-based rewards, achieving domain-general reasoning from symbolic math to creative poetry.

Microsoft Phi-2 a Small Language Model achieving high accuracy on scientific MCQ answering

In this 10-week project, groups of three aimed to enhance an existing large language model (LLM) to improve its performance in scientific question answering, specifically to evaluate its capabilities using EPFL exam questions. The model underwent fine-tuning and reinforcement learning with the Direct Policy Optimization (DPO) method and expanded its knowledge through Retrieval-Augmented Generation (RAG). To facilitate training on a single GPU, the model was quantized.

TripTracker

TripTracker is a mobile application designed to enhance the experience of exploring a city on foot by providing pleasant walking itineraries around interesting spots. Record and share your own walking paths, discover paths created by others, connect with your friends, and stay updated with trending routes. It's the ultimate travel companion that turns every trip into a memorable journey shared with the community! Group of 7 persons and coded in Kotlin.

Read more ⤢See Project →

Mysteries of luminara

The goal of this project was to develop a game for the Nintendo DS, an embedded system known for its significant limitations in memory, speed, interrupts, actions, and wireless connectivity. The game was created in groups of two and coded in C. This is a maze game where two players compete to reach the end without falling into holes. Each player can activate a potion to view the maze's endpoint and can see the other player's position in real-time.

Organising the mess in online news comments

This is my Bachelor project carried out at the LSIR laboratory (EPFL) under the supervision of Prof. Karl Aberer. The aim was to develop a web platform for summarizing user comments under newspaper articles with the help of finetuned neural networks.

Cellulo-Pong

This is a 14 weeks Unity-engine / C# project done in collaboration with 2 other people. The goal was to create a fun game using the Cellulos built by the CHILI-LAB of EPFL.

Read more ⤢See Project →

CryptKVS

This project was conducted as part of the CS-210 course. It is a project exclusively coded in C, and it was thought for groups of 2 students who were beginning to learn this language. This project covers most of the notions of C development.

Read more ⤢See Project →