2017 - (2020)
Arts & Technologies de l'Image, Université Paris 8 (Paris)
Computer Graphics, Computer Arts, 3D Modeling, Rendering, Compositing, Game Engines, Animation, VFX
2016 - 2017
M.Sc. in Computer Science — M2 MVA, École normale supérieure de Cachan (Cachan)
Applied Mathematics, Machine Learning, Graphical Models, Reinforcement Learning, Computer Vision, Medical Imaging, Computational Photography, 3D Point Clouds
2014 - 2015
M.Sc. in Computer Science — M1 MPRI, École normale supérieure (Paris)
Category theory, Computer Vision, Machine learning, Robotics, Software engineering, Quantum Computing.
2013 - 2014
B.Sc. in Computer Science, École normale supérieure (Paris)
Algorithmics, Compilation, Formal languages, Lambda calculus, Hardware systems, Operating systems, Networks, Signal processing. Passed with Mention Très Bien (highest honors)
2011 - 2013
Classes Préparatoires aux Grandes Écoles (MPSI, MP*), Lycée Saint-Louis (Paris)
Admitted at the École normale supérieure in Mathematics, Physics and Computer Science, ranked 20 (out of 1480)
Scientific Baccalauréat, Lycée Les Pierres Vives (Carrières sur Seine)
With music minor, Passed with Mention Très Bien
Apr. 2017 - Aug. 2017
Research intern in Computer Graphics, Télécom ParisTech, Paris, France
Geometry-Material-Lighting synchronized models for multi-resolution real-time rendering, supervised by Pr. Tamy Boubekeur.
Feb. 2016 - Aug. 2016
Software Developer, rise|fx, Berlin, Germany
Development of pipeline tools for Visual Effects artists.
Nov. 2015 - Feb. 2016
Independent Contractor in Deep Learning R&D, Interactions, Tele-working
Assist with Deep Neural
Networks optimization activities for Automatic
Development of a flexible experimentation setup.
May - Oct. 2015
Research intern in Deep Learning, Interactions, New York, US
Exploration of applications of Deep and Recurrent Neural Networks to Automatic Speech Recognition, supervised by Dr. Patrick Haffner.
June - July 2014
Research intern in Computer Graphics, IMAGINE, Inria Grenoble, France
Procedural generation of terrain from simple vector map using plate tectonics and erosion simulation on GPU, supervised by Pr. Marie-Paule Cani.
2014 - 2015
Developer and designer of CitizenWatt, http://citizenwatt.paris
Electrical consumption sensor and easy-to-use data visualization interface, supported by Paris city hall.
Python, C, C++, C#, Lua, OCaml, Java, MATLAB.
Various programming paradigms and software architecture.
Neural Net architectures (DNN, RNN, CNN).
Deep Learning tools (Torch, Theano, TensorFlow).
UI Design, Vector and pixel graphics, 3D modeling.
French (mother tongue), English (business level), German (ein bischen).
Harpsichord (5 yrs), Guitar (2 yrs), Piano (2
Studied solfège, and a bit of Music History.
Generation of Folded Terrains from Simple Vector Maps
Using Deep Neural Networks for Automated Speech Recognition
Élie Michel. M.Sc. first year internship report. 2015. [PDF]
Génération procédurale de paysages à partir de cartes vectorielles
Élie Michel. B.Sc. internship report. 2014. [PDF]
Jan 2015 (To be reviewed)
I have got an initial education centered on mathematics and physics, accompanied by knowledge computer science which I first acquired in my spare time and later through formal education at the École normale supérieure (Paris).
And I believe I could summarize my motivation by the fact that I like building.
I like building mathematical proofs out of axioms, logic circuits out of logic gates, programs out of instructions. And then I like combining all of them into more complexe systems and their multiple layers (Hardware, OS, Compilation, Program, Library, etc.) and their interactions (protocols, APIs). I also like discovering new building blocks (FPGA, Quantum computing).
This interest for building includes not only the result, but also the construction process itself, the way we need to design digital systems and organize their development and evolution.
Like any process, the architecture itself can be automated. It is called procedural generation and fascinates me because it is like teaching the computer how to build by itself. I have used it especially in computer graphics — particularly during my internship with M.-P. Cani — but we can also find auto-generated content, and even programs (genetic programming, machine learning).
But this automation still requires to communicate with humans, for them to explain what they need, and this issue meet another of my interests which is human/machine interaction. The machine should be able to understand the human.
And the opposite is interesting too. This interaction should provide humans both the ability to do things and the ability to do them easily. I like thinking about who is going to use the resulting product, how, and what for and design the experience and play with its affordances to build specific usages.
UI design also interests me for its aesthetic purpose. I have been playing with computer graphics and digital art since my first use of a computer and enjoy the issues, both artistic and technical, raised by matricial and vectorial drawing, 3D modeling and rendering, etc.
Still about getting the computer understand the human, I also investigated natural language processing, mixing formal grammar with the challenging human factor and the language reflexivity.
I think this human factor also motivates me to work with people with different backgrounds, and different points of view over what I do. I like teaching because I see it as yet another kind of construction, with human building blocks, and as for procedural generation, I enjoy seeing the creation growing by itself, learning from what I created myself.
I've once been ask for a visual summary of my UI design pieces of work.
CitizenWatt is a electrical consumption sensor that communicate though a web interface.
Freeder is a feed reader, and so has been focused on ease of reading and smoothness.
Evernest is a project of cloud event store service.
CitizenAir gathers geolocalized data from air quality sensors. Here is a poster presenting the project I made.
A walkthrough of levels 1 and 2 is available on YouTube.