Grégory R.

Data Scientist

660 dollar
9 years

My experience


Self-employedJuly 2017 - Present

I offer my skills to innovative companies entering or already present in the field of AI.

I do data science (machine learning and data analysis), computer vision, signal processing and robotics.


TAG HeuerOctober 2020 - Present

Client : TAG Heuer Connected, structure agile au sein de TAG Heuer, filliale du groupe LVMH.

Analyse et traitement du signal sur des données capteurs de la montre connectée de TAG Heuer. Directement rattaché au CTO.


EOS imagingNovember 2019 - June 2020

Client: EOS imaging is a global medical device company, leader in low-dose 2D and 3D medical imaging, that develops and markets advanced imaging and image-based solutions for musculoskeletal pathologies and orthopedic surgical care.
  • Develop 3D skeleton reconstruction software from biplanar X-ray images. 
  • Put machine learning models and image analysis algorithms into production (deep learning, statistical shape model, image registration, 3D deformation).
  • Normalise and improve methods and tools for image analysis R&D.

PhoventureJuly 2019 - October 2019

Client: Phoventure (solar plant investment and management company)

Build an ETL pipeline for solar irradiation data, and develop an automatic production and performance report generator.

  • Write a script to import solar irradiation data into the central database, set it up for scheduled automatic execution
  • Develop data aggregation and KPI computation functions
  • Develop an Excel report generator with Openpyxl
  • Integrate reporting generator with the management system interface

Hease GroupOctober 2018 - April 2019

Client: Hease is an innovating engineering and technological consulting firm specialized in robotics, AI and IoT.

Develop robot localization and obstacle detection algorithms as well as kinematics and sensor simulation, for the Heasy robot :

  • Implement a probabilistic mapping algorithm of moving obstacles 
  • Prototype and implement bluetooth beacon based localisation
  • Integrate a SLAM library (Simultaneous Localisation and Mapping)
  • Examine et correct IMU data to limit the odometry drift
  • Refine the kinematics and sensor probabilistic simulation
  • Add a moving obstacle scripted simulation feature to the simulator
  • Maintain and clean up the code base
  • Bonus : Optimise image streaming

StagoApril 2018 - September 2018

Client: Stago Group, a major player in the field of hemostasis and thrombosis

Implement support for a new hardware version of a blood analysis automaton (C#, C++).

  • Audit existing code and software architecture
  • Write and integrate C# / C++ classes allowing to handle a new robotic arm
  • Adapt the PLC code for the arm's operations

Spoon.aiDecember 2017 - April 2018

Client: Spoon, interactive robotics startup (artifical creatures)

Rebuild the Automatic Speech Recognition pipeline of the Spoony robot (C++, C# / Unity)

  • Benchmark Automatic Speech Recognition (ASR) engines.
  • Design the ASR pipeline.
  • Develop an Automatic Speech Recognition library in C++.
  • Integrate the C++ library in the robot's C# Unity code.

PhoventureOctober 2017 - November 2017

Client: Phoventure (solar plant investment and management company)

Build a full solar plant fleet data management and billing system from scratch.

Stack: Django (Python 3), MySQL, HTML, CSS, JQuery

  • Define and write specifications with the client.
  • Choose the stack and the cloud hosting solution.
  • Design the database model.
  • Develop management system (backend and frontend).
  • Write scripts to automatically import data from Excel sheets.

StagoJuly 2017 - September 2017

Client: Stago Group, a major player in the field of hemostasis and thrombosis (biomedical industry)

Demonstrate feasibility of using machine learning for anomaly detection in measurement data from a blood analyser.

  • Explore data, perform statistical analysis and visualisations.
  • Build and optimise a one-class SVM baseline.
  • Evaluate anomaly detection through autoencoding Convolutional Neural Networks (CNN).

Aldebaran - Softbank GroupDecember 2016 - May 2017

Softbank is the world leader in humanoïd robots.

Lead a team of 6 audio developers with the following objectives:

  • Enhance speech recognition disruptively with a PoC of a complete remold of the speech recognition signal processing pipeline
  • Respond to business needs regarding supported languages

Team achievements:

  • Improve reliability of speech recognition performance testing
  • Normalise test methods and metrics across the company
  • Allow adding new languages without software modification
  • Recast the audio signal processing pipeline of speech recognition
  • Synchronise with audio preprocessing solution providers in terms of requirements and schedule
  • Evaluate providers based on technical tests and conformity matrix
  • Continuously stabilise software

Aldebaran - Softbank GroupJuly 2015 - November 2016

Lead a team of 5 audio/vision developers with the following objectives:

  • Aid development of hardware version responding to software requirements
  • Adapt the algorithms to the future hardware

Team achievements:

  • Setup performance testing (speech recognition, people detection)
  • Put stereocamera calibration procedure in place
  • Adapt people detection algorithms for new sensors
  • Benchmark speech recognition engines in the cloud
  • Continuously stabilise software
  • Study legal and technical feasibility of audio data collection

Aldebaran - Softbank GroupAugust 2014 - July 2015

Scrum Master then Product Owner of a crossfunctional Scrum team focused on interaction starting and dialogue with the robot.

Stake: Allow the user to start an interaction easily and talk naturally with the robot. 

Conduct and contribute to following developments: 

  • Detection of keywords / interaction commands outside of any application
  • PoC of a hand wave detector
  • PoC of barge-in speech interruption by the human

Aldebaran - Softbank GroupMarch 2012 - July 2014

Develop the Pepper robot's listening capabilities to allow users to talk with Pepper fluidly.
Responsible for the speech recognition feature, technical and commercial relationship with the solution provider.

  • Develop and integrate an embedded sound source localisation algorithm (patent pending)
  • Evaluate and integrate and emotion recognition engine from voice
  • Make a PoC of Statistical Language Models (SLM) based on trigrams for speech recognition, then implement it
  • Prepare and stabilise several prototype demonstrations for investors in Japan as part of taskforces sent on site
  • Improve speech recognition latency and accuracy
  • Maintain the speech recognition module code: refactoring, stabilisation
  • Explore new potential partnerships

Involved skills:
Signal processing, Automatic Speech Recognition (ASR), Natural Language Processing (NLP), C++, CMake, Python, Numpy, Matplotlib, Git, Linux

My stack


Embedded Systems


Software Development, Machine Learning, Matplotlib, NumPy


Digital Signal Processing

Business Intelligence


Analysis methods and tools



Data analysis, Artificial Intelligence, Project Management, Algorithms, Research and development, Natural Language Processing (NLP), IoT, Data Science, Robotics



Environment of Development


Computer Tools

Microsoft Excel


Java, Python, C#, HTML, CSS, C/C++, C++

IT Infrastructure

Git, Linux


OpenCV, Django, jQuery

My education and trainings

Master of Science and Executive Engineering - minors : Applied Mathematics, Robotics, Computer Vision, Control Theory - National School of Mines of Paris2009 - 2012