Théophile Cantelobre

Machine Learning Engineer and PhD.

small.jpg

theophiledotcantelobreatinriadotfr

My name is Théophile, I live in Paris, and I am interested in technology.

I am a machine learning PhD (Inria, École Normale Supérieure) and engineer (Mines Paris, Sorbonne Univerité).

Open to work: I am actively looking for a Machine Learning Engineer position. Some things I value: a team to continue to learn with, an impactful product to build, and the opportunity to have an impact. I am open to branching out from ML to signal processing, optimization, systems engineering,… please get in touch!

I obtained my PhD in 2024 after three years in the SIERRA team at Inria Paris, working under the supervision of Alessandro Rudi, Benjamin Guedj and Carlo Ciliberto. During my PhD I developed and studied new methods in ML and statistical inference, with a focus on leveraging problems’ structure (such as invariances or constraints). I used a lot of (approximate) kernel methods and Pytorch and JAX on the implementation side. I worked on three types of data: images, time-series and filtering for dynamical systems.

I am passionate about building and running communities. I co-run the Mines Paris Alumni network with Philippe and the other volunteers. I also co-run the Mines Alumni Intelligence Artificielle group inside of MPA with Sebastián and Emma.

news

Oct 16, 2024 Successfully defended my PhD at Inria Paris!
Aug 16, 2024 Finished writing my PhD. My defense will be on the October 16th at 14:00 at Inria Paris.
Jun 18, 2024 Co-organized the first Mines Alumni Intelligence Artificielle seminar with Emma Bou Hanna and Sebastián Parratrieu with Félix Marty and Jingya Huang from Hugging Face. If you are in the Mines Paris network and would like to present, get in touch!
Apr 02, 2024 Presented Diffy at Imaging in Paris at IHP. Slides are here.
Feb 09, 2024 New preprint with Carlo, Benjamin and Alessandro is online!

selected publications

  1. Closed-form Filtering for Non-linear Systems
    Théophile Cantelobre , Carlo Ciliberto , Benjamin Guedj , and 1 more author
    arXiv 2402.09796, 2024
  2. Measuring dissimilarity with diffeomorphism invariance
    Théophile Cantelobre , Carlo Ciliberto , Benjamin Guedj , and 1 more author
    In Proceedings of the 39th International Conference on Machine Learning , 17–23 jul 2022
  3. A PAC-Bayesian Perspective on Structured Prediction with Implicit Loss Embeddings
    Théophile Cantelobre , Benjamin Guedj , María Pérez-Ortiz , and 1 more author
    arXiv 2012.03780,, 17–23 jul 2020
  4. A real-time unscented Kalman filter on manifolds for challenging AUV navigation
    Théophile Cantelobre , Clément Chahbazian , Arnaud Croux , and 1 more author
    In 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 17–23 jul 2020