Description

I am a 27 years old french data scientist.

Having a Data Science background I enjoy working on real projects that matters. My previous experiences in famous companies such as L'Oréal and Hewlett Packard gave me the opportunity of being involved in all the parts of a Data project. I took part many times in data collection and data cleaning during the past years. I also had the chance to design sophisticated algorithms and put them in production through cloud inference pipeline and iOS integration using Swift and CoreML. I even had the great pleasure to work on a paper presented at ECCV 2020 and be named in the acknowledgments : AgingMapGAN.

I deeply appreciate working in this field which is in constant evolution. All projects have their subtleties and that's why I focus mainly on delivering key outcomes and quickly adapting to new learnings.

Technical Stack:

• DataScience : Tensorflow (Keras), Pytorch, Scikit-learn, Pandas, Numpy, Matplotlib

• Cloud : GCP/AWS/Azure

• WorkFlow : Git, Docker, CookieCutter,

• Language : Python, Django, SQL, Swift, JavaScript

• Visualisation : Tableau, PowerBI, Flask

Languages

French
Native or bilingual
English
Fluent
Chinese
Basic

Workplace preferences

Can work on-site

Paris (up to 10km)

L'Oréal - Innovation
Data Engineer
FASHION AND COSMETICS
May 2021 - Today (5 years and 1 month)
Paris, France
L’Oréal is a French industrial group of cosmetic products. The company has nowadays become an international group and is one of the world leader in cosmetics industry.

Project & Tasks :

• During this second mission at L'Oréal I am in charge of integrating the previously made "face analysis"** pipeline into an Ipad. I am working on iOS with Swift to integrate the python based AI algorithm using CoreML in order to optimize performances. This is to be able to run the diagnostic directly on device, without internet access, which also provides more security knowing that the data doesn't need to go through the network.

• I do some refactoring on the models to adapt their inputs and outputs type to get the most out of the CoreML framework and swift pipeline.
In addition I work to reduce their size and their memory needs in order to be able download them more easily on device and doing inference with multiple models in parallel within the Ipad.

• To have tracking and maintainability we did a CI/CD pipeline using Github Actions to automate the testing and delivery.

• I also use Objective-C and C++ in order to do some image processing on device.

** Details of the first mission below
Machine learning Deep Learning Swift TensorFlow Pytorch iOS Python Xcode Scikit-learn Git Pandas
SNCF RESEAU
Data Scientist / Data Engineer
TRANSPORTATION
July 2020 - April 2021 (9 months)
Saint-Denis, France
SNCF is the French public railway company. It is particularly present in the fields of passenger transport, freight transport and manages, operates and maintains the national rail network belonging to the State.

Projects & Tasks:

• Being part of the data team as a deep learning consultant, my work focuses on the rail network analysis and delay prediction.

• I worked on some Django app in order to provide internal endpoints.

• I also participate on some data engineering tasks such as the models testing and deployment. For that matter I was in charge of creating a cookicutter template which helps to train, test and deploy data projects. The idea was to embedded the training and inference code inside docker images which can then be used in AWS SageMaker with S3. We also used AirFlow and Gitlab-CI/CD.
Data Engineer Deep Learning Machine learning Amazon Web Services TensorFlow Docker Django Python Git flask
L'Oréal - Innovation
Data Scientist / Data Engineer
FASHION AND COSMETICS
July 2019 - June 2020 (11 months)
Paris, France
L’Oréal is a French industrial group of cosmetic products. The company has nowadays become an international group and is one of the world leader in cosmetics industry.

Projects & Tasks:

• Part of the AI team as a deep learning consultant I worked on computer vision tasks for specific facial features such as wrinkles. Most of the models went in production for the Lancôme project (link below)

• I also participate on the writing of the paper AMGAN which is linked below. I focused on the GAN training and architecture alongside Julien Despois. The aim was to change the age and some other facial features on portrait pictures. We demonstrate the advantage of our proposed method in terms of quality, control, and how it can be used on high-definition images while limiting the computational overhead. The project was presented at the ECCV 2020.

• We also tackled the explainability angle in order to monitor the algorithms training and performances with increased insights. Some visualizations were made in order for the client to see on which pixels the model based its score.

• I also worked on some NLP tasks such as social network's text processing in order to analyse the feedback (sentiment, similarity, top topics, knowledge graph) and use the customers opinion
Deep Learning Machine learning TensorFlow Pytorch Python flask Git