Welcome, Bienvenue !
I am a Machine Learning Researcher in Apple's MLR team, and I am based in Copenhagen.
I recently completed a PhD student in Machine Learning and Cognitive Science at the Ecole Normale Supérieure, under the supervision of Emmanuel Dupoux and Guillaume Wisniewski. I was also an intern at Meta as an AI Research Scientist Intern, in the Paris office.
Previously, I graduated from the University of Edinburgh with an MSc in Speech and Language Processing, and from City, University of London with a BSc in Psychology. I also worked as a speech recognition engineer, focusing on acoustic modelling, prior to starting my PhD (see my full CV here).
Research Interests
My recent research focused on bilingual language acquisition, from a reverse-engineering perspective. Briefly, it consists in constructing unsupervised models of cognitive behaviours, using the same input available to humans, as a way to draw hypotheses on the actual cognitive processes at stake. More information on this approach is described on the CoML team’s website.
I am a strong advocate for the importance of tying links between the Machine Learning and Psychology communities, as I believe both benefits from getting insights from the other.
I am also currently a part of the committee for the 2021 Zero Resource Speech challenge, which focuses on evaluating unsupervised speech representation models. Check it out!
Latest news
26/08/2024 - [ 💼 WORK] I started a new position at Apple MLR as a Machine Learning Researcher.
15/08/2024 - [📄 THESIS] My PhD thesis is now available online.
13/01/2024 - [📄 PAPER] We just released EmphAssess, a result of my work at Meta, which is a benchmark aimed at assessing emphasis in speech-to-speech models. The related preprint is available in ArXiv.
27/11/2023 - [ 🎓 PhD defense] I successfully defended my PhD defense at the Ecole Normale Supérieure!
26/06/2023 - [ 💼 INTERNSHIP] I just started a 5 months internship at Meta in the Paris AI research group.
29/05/2023 - [📄 PAPER] Our new paper "Realistic and broad-scope learning simulations: first results and challenges" has just been published in Journal of Child Language! We discuss the future of learning simulations for early language acquisition.
24/02/2023 - [🔎 BENCHMARK] + [📄 PAPER] We just released ProsAudit, a benchmark for evaluating English prosody in SSL models of speech! It is now part of the Zero-Resource speech challenge, and the related preprint is available on ArXiv.