Rémy Frenoy

Work Experience

• Jellysmack - Lead AI

  Nov. 2021 - present

  • Jellysmack is the global creator company that detects and develops the world's most talented video creators through technology.
  • Leading a team of 10 data scientists, we make products that help video editors.
  • Leveraged knowledge: Image/video processing, Computer Vision.

• Octo Technology - Senior Machine Learning Engineer

  Feb. 2019 - Nov. 2021

  • In addition to my work as an engineer, I am supervising and mentoring teams of data scientists and ML engineers to be up to date with state of the art solutions and apply software development best practices . I have worked on projects using the most common cloud service providers such as AWS, GCP or Azure. Examples of projects: Creating a predictive maintenance software for a major actor in the energy industry, creating a software that detects bacteria from pictures of petri dishes for an actor of the pharmaceutical industry.
  • Leveraged knowledge: Same as below, plus Knowledge sharing, Cloud environments (AWS, GCP, Azure).
  • Open source contribution: psquare, implementation of The \(p^2\) algorithm for dynamic calculation of quantiles and histograms without storing observations by Jain and Chlamtac.

• Octo Technology - Machine Learning Engineer

  Feb. 2017 - Nov. 2021

  • Help tackling business issues with software and data (ML involved or not).
  • Working on ways to improve our development/deployment tools and methods to make it easier/faster/safer to deploy efficient machine learning and data pipelines in a production environment.
  • I am also a trainer for the Advanced Data Science courses (levels I and II) and Apache Spark courses given by Octo Academy.
  • Leveraged knowledge: Python programming (pandas, pyspark, numpy/scipy, scikit-learn, tensorflow, pytorch, unittest/pytest, setuptools), Airflow, Jenkins/Gitlab CI. Exploratory Data Analysis. Supervised learning (mostly ensemble models on tabular data, time series). Model selection. Object-Oriented and Functional programming. Agile organization.

• University of Technology of Compiègne - PhD Student

  Oct. 2013 - Oct. 2016

  • Topic: User modeling and adaptation of action selection in interactive training environments
    • Error recognition: Studied and used supervised learning models in order to recognize patterns of error from users gestures (supervised multilabel classification).
    • User modeling: Studied and developed a framework where users are represented in a multidimensional metric space, making it possible to extract a personalized learning path and monitor the evolution of users' profiles over time.
    • Adaptation: From user modeling, adaptation of action selection can be seen as a multi-armed bandit problem. Proposed and implemented a contextual multi-armed bandit algorithm to make a system automatically and dynamically adapt to the user over time. This work has been published as a full paper at the ACM User Modeling, Adaptation and Personalization Conference (ACM UMAP 2016).
  • Leveraged knowledge: Supervised learning, Bandit algorithms, Python programming, Version control (git), Technical writing.
  • Publications:
    • Adaptive Training Environment without Prior Knowledge: Modeling Feedback Selection as a Multi-armed Bandit Problem
    • Contributions of mixed reality in a calligraphy learning task: Effects of supplementary visual feedback and expertise on cognitive load, user experience and gestural performance
    • (In French) Modélisation de l'activité gestuelle et sélection automatique de feedback pour des environnements interactifs d'apprentissage : application à la calligraphie
  • Teaching activities:
    • Artificial Intelligence course (IA04): teaching assistant.
    • Introduction to virtual reality (RV01): teaching assistant.
    • Participated in the creation of the practice exercises (french content).

• Praxedo - Software Engineer

  Feb. 2013 - Oct. 2013

  • Built prototypes based on several NoSQL solutions (especially MongoDB) to measure their efficiency/scalability within the context of Praxedo data. Proposed a new version of MongoDB data balancer to optimize update actions on sharded clusters.
  • Leveraged knowledge: Programming in Java/Javascript, knowledge of advantages and limitations of different kinds of data storage paradigms, deploying/monitoring/optimizing MongoDB reading/writing operations under various situations.
  • Publication: Design and Implementation of a MongoDB solution on a Software As a Service Platform