About Me

Hi! I'm a first year ELLIS PhD student in mechanistic interpretability and AI safety at the MPI for Intelligent Systems advised by Prof. Bernhard Schölkopf, and supported by the Hector Fellow Academy.

I am interested in understanding the internal mechanisms of large language models (LLMs), particularly their "world models"—the structured representations they develop to interpret and make sense of data. My current projects center around cross-layer transcoders and attribution graphs.

Before starting my PhD, I did a BSc in mathematics at EPFL and spent one year at ENS in Paris (Ulm). I then transitioned to pursue a MSc in Neural Systems and Computation at ETH Zürich. During my master, I worked by Prof. Maneesh Sahani at the Gatbsy Unit in London on non-generative representation learning.

Profile Picture

Florent Draye

PhD Student in Machine Learning

MPI-IS, Germany

Education

  • PhD, Machine Learning

    Max Planck Institute | 2024–Present

  • MSc, Neural Systems and Computation

    ETH Zürich | 2022–2024

    GPA: 5.9/6

  • Exchange Program, Mathematics

    Ecole Normale Supérieure (Ulm) | 2021–2022

    GPA: 18.3/20

  • BSc, Mathematics

    EPFL | 2019–2022

    GPA: 5.88/6

Prizes

  • Best Bachelor Average Grade (1500$)

    Awarded for achieving the highest bachelor average (5.88/6) among 1214 graduates, 2022.

    EPFL, 2019–2022

  • Part III International Scholarship – University of Cambridge

    Awarded based on academic potential for the Master of Applied Mathematics (Part III), declined.

    University of Cambridge, 2022

Experiences

  • Machine Learning Research Intern

    Gatsby Unit UCL, 09/2023-02/2023

  • Data Science Intern

    Mindmaze, Summer 2023

News

Just started my PhD with Prof. Bernhard Schölkopf

January 2025

Research

A Review on the Universal Approximation Theorems of Artificial Neural Networks

U Faure, F Draye, Bachelor Thesis 2022

† Equal contribution

PDF