I am Theo Lepage, a Ph.D. student in Artificial Intelligence at Sorbonne University and EPITA Research Laboratory, based in Paris, France. My research focuses on self-supervised learning frameworks to develop robust representations for speaker and language recognition. Through machine learning and speech processing, my aim is to enhance communications between humans and computers.
Read my resume
Send me an email at theo@theolepage.com or reach me on 
.
Publications
Exploring WavLM Back-ends for Speech Spoofing and Deepfake Detection
ASVspoof 2024
Theophile Stourbe, Victor Miara, Theo Lepage, and Reda Dehak
Towards Supervised Performance on Speaker Verification with Self-Supervised Learning by Leveraging Large-Scale ASR Models
INTERSPEECH 2024
Victor Miara, Theo Lepage, and Reda Dehak
Additive Margin in Contrastive Self-Supervised Frameworks to Learn Discriminative Speaker Representations
Odyssey 2024
Theo Lepage, and Reda Dehak
Experimenting with Additive Margins for Contrastive Self-Supervised Speaker Verification
INTERSPEECH 2023
Theo Lepage, and Reda Dehak
Label-Efficient Self-Supervised Speaker Verification With Information Maximization and Contrastive Learning
INTERSPEECH 2022
Theo Lepage, and Reda Dehak
Talks
Self-Supervised Learning for Speaker Recognition
Dec. 2024 @ Callyope — Paris, France
Additive Margin in Contrastive Self-Supervised Frameworks to Learn Discriminative Speaker Representations
Jun. 2024 @ Odyssey — Quebec, Canada
Experimenting with Additive Margins for Contrastive Self-Supervised Speaker Verification
Aug. 2023 @ INTERSPEECH — Dublin, Ireland
Label-Efficient Self-Supervised Speaker Verification With Information Maximization and Contrastive Learning
Sep. 2022 @ INTERSPEECH — Incheon, South Korea
Lightning Talks on SSL applied to Speaker and Language Recognition
Jan. - Sep. 2021 @ LRDE — Paris, France
Teaching
Introduction to Deep Neural Networks
Spring 2023 - 2024 @ EPITA
Python for Data Science
Spring 2023 - 2024 @ EPITA
Rust Programming
Spring 2020 @ EPITA
Unix / C Programming
Fall 2019 @ EPITA
Projects
sslsv
Framework for training and evaluating self-supervised learning methods for speaker verification.
wavlm_ssl_sv
SOTA method for self-supervised speaker verification leveraging a large-scale pretrained ASR model.
tensorflood
Tiny automatic differentiation (autodiff) engine for NumPy tensors implemented in Python.
© 2025 Theo Lepage