Guillaume Salha-Galvan

Hi! I am Guillaume, from France. I am an incoming Associate Professor in Information Engineering at Shanghai Jiao Tong University (SJTU). Starting this summer, I will join the SJTU Paris Elite Institute of Technology, an international research and teaching institute established at SJTU in partnership with four French engineering schools: École Polytechnique, Mines Paris, Télécom Paris, and ENSTA.

Before joining SJTU, I gained nearly a decade of industry experience, including as Director of Research at the sustainability-driven organization Kibo Ryoku and as Research Scientist and Coordinator for Music Recommendation at Deezer, a leading French music streaming service. You can find details about my experience on LinkedIn.

My research interests span the broad areas of machine/deep learning, graph learning, recommender systems, large language models, and their applications to music. I have published over 30 scientific articles in peer-reviewed journals and leading conferences, including ICML, KDD, WWW, IJCAI, EMNLP, and RecSys. I have received best paper awards or honorable mentions at RecSys (2020, 2021, 2024) and ECIR (2024). Several of these contributions have had real-world impact, notably by powering Deezer's recommender systems, which help millions discover music.

I completed my Ph.D. in Computer Science at École Polytechnique, IP Paris. My doctoral research, supervised by Romain Hennequin and Michalis Vazirgiannis, was conducted in parallel with my work at Deezer, through a scientific and industrial partnership between both institutions. Prior to that, I graduated from ENS Paris-Saclay with an M.Sc. in Machine Learning (MVA), and from ENSAE Paris with an M.Eng. in Data Science, both with first-class honours.

I am always looking for professionals with common interests to learn and share experience, so feel free to reach out!

News

May 2025

Our paper on von Mises-Fisher sampling has been accepted at ICML 2025!

Mar 2025

Thrilled to co-organize the EARL workshop on evaluating and applying recommender systems with LLMs. Happening this September in Prague during RecSys 2025!

Feb 2025

We will present our study on vMF sampling of GloVe vectors at the FPI@ICLR 2025 workshop.

Jan 2025

Our paper on weight sharing in variational graph autoencoders has been accepted at WWW 2025.

Oct 2024

We are honored to be among the "Best Full Paper" candidates at RecSys 2024.