Fabien Cellier (PhD)

Researcher, Entrepreneur, Data Scientist...

CV Français

Basic Information

Latest Work Experience


Founder / CTO - DaSciLs SASU (October 2015 - 2018)

  • Creation of Environmental and Session based Recommendation/Prediction of estates for HomeInLove
    -> Big Data innovation award from Arcelor-Mittal 2016 - Deep Learning and Research
    -> Reward of BPIFrance for technological innovation (i-lab 2016)
  • Web site/application with features based on geolocation
  • Cloud infrastructure for HPC processing (GPGPU, HDInsight...)
  • Recruit and Manage IT team for HomeInLove
  • Supervise the collaboration with research laboratories ( liris and centre-max-weber)
  • Data Ingeneer (Société général) : synchronization between Databases and the datalake

External Teacher - ESCPE Lyon (September 2014 - Present)

  • Teach WebGL and introduction to Data Science and Big Data

DataScientist - Worldline (March 2015 - September 2015)

  • Integrate and analyze data to use it in "data value factory":
    big data platform to share user information and launch advertising campaigns on specific target
  • Creation of automatic tools for other data scientists based on spark/python/R:
    statistical reports, unsupervised clustering, supervised classification with prediction
  • 15% of my time on research thematic:
    Deep learning for Prescriptive analysis (Q-learning/SARSA) and add prediction to RFM (recency, frequency, monetary)

Member of Khronos as a consultant - Khronos (January 2012 - December 2015)

  • Member of Khronos as a consultant in the WebCL Working Group

Previous Experience

Research engineer - Worldline (December 2012 - March 2015)

  • Adaptation of a PaaS, OpenShift/node.js and HAProxy, to be able to use GPGPU without introducing new security threats.
  • Support in creating research projects and collaboration with other firms on HPC based on stream integration with spark(hadoop/mesos and cassandra for storage)
  • Distributed computing with low latency (GPGPU as a service)
  • Trans-compiler from GPU code to secured Code with emulation for fallback

PhD student/ Engineer - Worldline (Gis department)/Liris CNRS (September 2009 - December 2012)

  • Research Domain : Geometrical Modeling, Computational Geometry, streaming and cloud computing
  • Creation of a 3D GIS client, Nginx module for the data, WebGL/WebWorker and WebSocket for the client. Basis for the new product OpenscalesGL
  • Creation of new algorithms for the streaming based on geometric with OpenMesh
  • Creation of a new method for 3D model simplification published in scientific paper and based on GPGPU with OpenCL 1.1
  • Development of a WebCL prototype without security in firefox with WebGL interaction in January 2011

Developer- Worldline (Gis department) (March 2009 - September 2009)

  • Conception and deployment of a shared platform for visualization of geographical data. Responsible of the creation of the security and authorization token in Atos version of GeoServer

Developer- Worldline (Gis department) (October 2007- October 2008)

  • Conception, creation and deployment of a website to administrate geoportail : visualization of geographical data and creation of the security and the reporting in the back office based on
    java5, spring2, struts, maven, dwr, GeoServer, OpenLayers, PostgreSQL/PostGIS.

Skills

Technologies

  • C/C++, Java, Python, OpenCL(WebCL), OpenGL, Scala, Javascript, WebGL, Typescript.
    In my spare time: Rust

Framework

  • Pandas, scikit-learn, Tensorflow, Keras (neural network), theano, chartjs, Spark, zeppelin, hapi,express, deeplearning4j

Algorithm

  • NoSQL, MapReduce, clustering, feature selection (PCA, fisher), classification, regression, model selection (boosting/stacking), Natural Language Processing (word2vec,lstm,rnn)
    Specialization (research) : Artificial Neural Network (cascade 2 with LSTM, mlp, deep stacking network, deepautoencoder) , 3D mesh simplification

Product and tools

  • centOS, Nodejs ,git, mercurial, vim (vundle), netbeans, Docker, hdfs, yarn,, kubernetes, Azure, haproxy, nginx,Mongodb,PostgreSQL/PostGIS, cassandra