Resume

Download (.pdf)

Theo Lepage

Speech/Voice AI Researcher•Ph.D. in AI•Self-Supervised Learning & Speaker Recognition

Paris, France (open to relocation EU/US)

contact@theolepage.com

Experience

Ph.D. Researcher at

Sorbonne Université / LRE-EPITA

Paris, France

Nov. 2022 - May 2026

Developed self-supervised models for speaker verification, with DINO-WavLM achieving SOTA on VoxCeleb (1.06% EER on Vox1-O)
Proposed SSPS, a latent-space positive sampling for SSL frameworks, mitigating extrinsic variability (-58% EER for SimCLR)
Built and maintained sslsv, an open-source PyTorch toolkit for training and evaluating self-supervised speaker models
Published 8 papers in leading speech venues including Interspeech, IEEE TASLP, and Speech Communication

Research Scientist (Internship) at

Siemens Healthineers

Princeton, USA

Feb. 2022 - Sep. 2022

Developed deep learning models (CNN with self-attention) for end-to-end MR image enhancement (denoising & super-resolution)

Software Engineer (Internship) at

Paris, France

Sep. 2020 - Jan. 2021

Contributed to Holovibes, real-time digital holography software for retinal blood flow analysis → 20× input throughput (10,000 FPS)

Education

Sorbonne Université (Ph.D. in Artificial Intelligence)

Paris, France

Nov. 2022 - May 2026

Thesis: Self-Supervised Learning for Speaker Recognition • Supervised by Reda Dehak and Thierry Géraud @ LRE-EPITA

École Pour l'Informatique et les Techniques Avancées - EPITA (M.Eng. in Computer Science)

Paris, France

Sep. 2017 - Sep. 2022

GPA: 3.9/4.0

Major: AI/ML for Computer Vision • Research student • Teaching assistant (C, Unix) • International section • Exchange semester @ CSUMB

Publications

Selected first-author articles • Full list on

Self-Supervised Learning for Speaker Recognition: A study and review

Speech Communication•2026

SSPS: Self-Supervised Positive Sampling for Robust Self-Supervised Speaker Verification

Interspeech•2025

Self-Supervised Frameworks for Speaker Verification via Bootstrapped Positive Sampling

IEEE TASLP•2025

Projects

speakerscope.ai

Speaker diarization with ID, gender, language, and emotion insights via web or API, powered by SOTA AI models.

Deep learning toolkit based on PyTorch for training & evaluating self-supervised models for speaker verification.

Self-supervised framework to fine-tune WavLM for speaker verification, without labels, achieving SOTA on VoxCeleb.

Miscellaneous

Skills

AI/ML:

PyTorch

TensorFlow

Scikit-learn

NumPy

Pandas

Programming:

Python

C

C++

CUDA

JavaScript

Bash

Tools:

Git

LaTex

Docker

Slurm

SQL

Languages:

English (fluent)

French (native)

Teaching

Intro to Deep Learning (2023-2025 @ EPITA)

Python for Data Science (2023-2025 @ EPITA)

Awards & Honors

3rd place @ ASVspoof 5 (Track 1, Open)

Academic Service

Reviewer for Interspeech & ICASSP