Antoine CAUTRU

Experience

• 

  Centrale 7

  Sept 2018 - Jun 2020

  Global Communications Manager
• 

  GUARDTEX

  Jun 2019 - Aug 2019

  Stagiaire

  Assistance with cutting and assembling materials for sailing equipment- using a cutter machine- using a laser cutter machine - sewing sails and boat protections

• 

  Capgemini

  Sept 2019 - May 2020

  Ambassadeur chez Capgemini

  Responsible for promoting the company on the campus of CentraleSupélec

• 

  Reuniwatt

  Aug 2020 - Feb 2021

  Data Scientist

  Geometric, physical, and statistical combinations of sky observation cameras to improve solar forecasting. We tried not to use Machine Learning to see how far we could go to discover relations between the following complex data. Using several sensors (pyranometers, ceilometers, sky cameras), we were able to challenge the state-of-the-art for the computation of:- the Cloud Based Height (CBH) with two different methods of stereoscopy (using images), validated by the ceilometer measurements- the speed and direction of the wind with two different methods (using images or pyranometers)- the short-term (30 minutes) power forecasting of a solar plant.The various challenges involved:- Object Oriented Programming (python) - Use of Pandas, Numpy, Matplotlib- 3D geometry for the projection of images at different heights until the two projection matched (stereoscopy theory)- 2D geometry to link the delay between the pyranometers' measurements and the speed and direction of the wind- data processing for image and signals filtering, for more robust and stable computing of complex metrics (such as the CBH) Show less

• 

  Milliman

  Mar 2021 - Aug 2021

  Data Scientist

  Studies on the contribution of public data in insurance.The aim of the project was to explore the potential of publicly available data in insurance, in order to optimize the contract with new clients. Technically, the role involved:- scrapping data from complex data bases (Google Street View, Google Sky Images, OpenStreetMap, Governmental data)- complex analyses of images to automize the detection of damage on street views- finding correlations between different data sources- predicting the risks for living in a particular area for car accidents, water damage, wildfires- etc. Show less

• 

  The University of British Columbia

  Oct 2023 - Mar 2024

  Master's thesis

  Assessment of pathological outcomes of Traumatic Brain Injury (TBI) in the presence of Alzheimer’s Disease (AD)In my master's thesis, I explored the reciprocal impact of Alzheimer's disease (AD) on traumatic brain injury (TBI) outcomes. Using AD mouse models, I investigated changes in neuronal activity and amyloid pathology following TBI. My work combined conventional histological methods with cutting-edge 3D brain-wide assessments using Light Sheet Microscopy (LSM). This research highlighted distinct mechanisms of neuronal activity around amyloid plaques and laid the groundwork for advanced 3D analysis techniques in neurodegenerative disease studies.The literature analysis explored:- the pathological outcomes with the CHIMERA model in terms of amyloid plaques and neuronal activity- a histopathological analysis of AD- stains for amyloid plaques - amyloid plaques staining, detection, and quantification in the whole brain.The coding challenges involved:- complex manipulation of images using morphological transformations for segmentation and filtering- machine learning (Random Forest) and deep learning (CNN) for segmentation of images- 2D and 3D points distribution analyses for the discovery and complex patterns linking amyloid plaques and altered neurons- proposing reproducable visualization tools of 3D microscopy images, by segmenting brain regions with the Allen Brain Atlas- using python (Oriented Object Programming, Pytorch, OpenCV, Numpy, Pandas)- parallelization of algorithms for heavy images (Multiprocessing library) Show less

• 

  Université de Montréal

  Apr 2024 - Dec 2024

  Research Intern

  Development of a python module for biologically relevant Spiking Neural Networks with PyTorch, at the Precision Psychatry and Social Physiology Laboratory.Inspired by the paper "Multilevel development of cognitive abilities in an artificial neural network" by Volzhenin, Changeux, and Dumas, this project replicates and extends the original work by introducing batch processing with PyTorch to explore cognitive epigenesis. We implemented bidirectional synapse strengthening, spatial neuron positioning, time delays, and distance-based synaptic conductance decay, leveraging GPU acceleration for faster, more efficient simulations. Additionally, we proposed a biologically relevant optimizer to accelerate learning while maintaining stability. Our module enables more complex STDP rules and improves upon BindsNET and Brian by achieving similar results with shorter simulation times.The technical challenges involved:- implementing STDP learning rules of neurons with matrix products in a Spiking Neural Network- finding optimal hyperparameters for the reproducibility of results obtained with Brian (which uses event-driven computing)- implementing delayed synaptic transmissions with matrix products, using replay buffers- implementation of a learning rule for the processing of the simultaneous weights updates across batches while maintaining network stability- visualization and qualtification of network performances Show less

• 

  Neurinnov

  Jan 2025 - now

  R&D engineer