Guillaume Dehaene

Artificial Intelligence Engineer

Professional experience

Artificial Intelligence Engineer

Geodaisics | Sept. 2023
  • Creation and implementation of the internal validation procedures of AI models.
  • Creation of a shared workflow and tooling for the research team.
  • Wrote and reviewed developpement procedures for the internal Quality Management System in accordance with ISO 13485 / 14971 / 62304 / …

Senior data Engineer

Marelli - Smart Me Up | July 2022 - July 2023
  • Project supervision: embedded gesture recognition (precision >95%).
  • Definition of data needs for the Grenoble teams.
  • Responsible for the internal online data annotation tool.
  • Supervision of the annotation team (12 people). Creation and supervision of the data sharing procedures.

R&D computer vision engineer

Marelli Smart Me Up | April 2020 - June 2022
  • Creation and implementation of an unsupervised stereo vision algorithm.
  • Creation of an internal library to standardize R&D activity on neural networks.
  • Features: automated code standards, unit testing, web visualization of results, reproducibility.
  • Technology watch on computer vision. Algorithms adapted: SwAV, Mean teacher, depth estimation, transformer.

Assistant professor in Statistics

Ecole Polytechnique Fédérale de Lausanne | Sept. 2016 - April 2020
  • Neurips 2016 AABI workshop Disney Research Paper Awards awarded for: Expectation Propagation performs a smoothed gradient descent, G. Dehaene.
  • Creation and implementation of a method to validate the results of a Bayesian statistical analysis
  • Supervision of one PhD. and three master theses.

Skills

Computer science

  • Python (expert): django, pytorch, tensorflow
  • Rust
  • Javascript
  • HTML, CSS
  • git
  • Docker
  • Linux admin

Management

  • Agile project management
  • R&D supervision
  • CI / CD

Artificial Intelligence

  • Neural networks
  • Computer vision
  • Statistical theory
  • Bayesian statistics

Languages

  • French (native)
  • English (Bilingual)

Education

Ph.D. in neuroscience and statistics
Université de Genève, Université Paris- Descartes
2012-2016

Ecole Polytechnique engineer diploma
Ecole Polytechnique Paris
2008-2012

Master in Cognitive Science
ENS-Paris, EHESS, Université Paris-Descartes
2011-2012

Publications

A deterministic and computable Bernstein-von Mises theorem
G. Dehaene, 2019
Presented at: Séminaire BIG (Grenoble), Séminaire de Statistique de Berne

Expectation Propagation in the large data limit
G. Dehaene and S. Barthelmé, 2018
Journal of the Royal Statistical Society - Series B
Presented at: Séminaire BIG (Grenoble), Séminaire de Statistique de Genève

Expectation Propagation performs a smoothed gradient descent
G. Dehaene, 2016
Advances in Approximate Bayesian Inference NeuRIPS workshop
NeurIPS AABI Workshop 2016 Disney Research Paper Awards

Bounding errors of Expectation-Propagation
G. Dehaene and S. Barthelmé, 2015
NeurIPS 2015

Hobbies

Back to top