four40.work

Neo Stories

DataCrafters

Agora

Copy House

Omnitech

Code6

MVP Rocket

Breakfast in America

Sigmoon 🌕

Digital Learning 4U

Air et Marine

MoBo Web Squad

Adia

Hussar

Maximedian

Byteworks

HeroGrowth

Social Clean - https://socialclean.fr/

<ul><li><p>Build NLP (CamemBERT) and Computer Vision (YOLO) models to analyze user networks.</p></li><li><p>Extraction of image and text data, analysis (categorization of comments such as drugs, alcohol, insults etc.) and "sensitive" images (presence of cigarettes, alcohol bottles etc.).</p></li><li><p>S3 and Big Query for data storage and processing, FastAPI and Python for interfacing, Github + Docker + Cloud run for deployment</p></li></ul><p></p>

Textual and visual analysis of user's social network, looking for content that could damage the user's e-reputation..

Python

Data science

Jules Desclous

I recommend Sylvain who is a top profile! It was a chance to have him in our team during our different projects.

He was able to adapt to our technical stack, he was able to work successfully on various projects using different technologies, showing his ability to adapt quickly to a new technical environment.

His integration into the team has always been easy thanks to his positive attitude and his ability to work in a team. He is always willing to share his knowledge and help the team.

Sylvain is also very resilient, able to handle challenges encountered in projects in an efficient manner, finding creative solutions to problems.

He also has a quick learning curve on new tools, which allows him to always stay up to date with the latest technologies.

Finally, Sylvain is able to manage a project from A to Z in an efficient manner, respecting deadlines and guaranteeing quality delivery.

--

Lucas Dahan :

Great work with Sylvain on a challenging MLOps mission :
- Build from scratch a complexe system of API matching lowcode front, social networks API, and Google Cloud Plateform 
- Build a full data pipeline on GCP moving from different predictions steps (data are unstructured)
- Automate a complete AI cycle on GCP (data collection, data transformation, prediction, restitution) based on custom trigger
- Create and deploy two successfull AI models for text analysis and object detection

NLP + Computer Vision: e-reputation B2C app

Airbus Defense and Space

<p><strong>Customer Need:</strong> Airbus Defense and Space, in collaboration with the European Space Agency (ESA), sought to ensure the reliability and effectiveness of the Jupiter Icy Moons Explorer (JUICE) Spacecraft's Thruster Tranquillisation Mode. In particular, the focus was on ensuring stability during various phases, including Earth fly-bys, and guaranteeing robust control across multiple scenarios.</p><p></p><p><strong>Derived Solution:</strong></p><ol><li><p><strong>Controller Design and Tuning</strong>: A multidisciplinary approach was employed, centering on optimal controller design and tuning. This encompassed nonlinear control law adjustments tailored to the JUICE Spacecraft.</p></li><li><p><strong>Optimal Sensor Fusion and State Estimation</strong>: Innovations were introduced to achieve a synergy of sensors and enhanced state estimation capabilities for improved spacecraft control.</p></li><li><p><strong>System Performance Verification</strong>: Monte-Carlo methods, known for their reliability in complex systems, were utilized to verify the performance of the JUICE Spacecraft systems.</p></li></ol><p></p><p><strong>Value Delivery:</strong></p><ul><li><p><strong>Enhanced Spacecraft Stability</strong>: Successfully identified and addressed a crucial non-convergence issue, ensuring the spacecraft’s stability when transitioning out of the Thruster Tranquillisation Mode.</p></li><li><p><strong>Boosted Confidence in Earth Fly-Bys</strong>: Introduced innovative 'robustness' test scenarios, heightening confidence in the spacecraft's capability to diminish potentially harmful structural modes during abnormal Earth fly-by conditions.</p></li><li><p><strong>Advanced Verification Framework</strong>: Elevated the Monte-Carlo frameworks, which is crucial for the statistical verification of intricate spacecraft control systems.</p></li></ul>

Ensured the robustness and reliability of the ESA's Jupiter Icy Moons Explorer (JUICE) Spacecraft's Thruster Tranquillisation Mode. Through optimal controller design, nonlinear control law tuning, advanced sensor fusion, and state estimation, critical issues like potential non-convergence during mode transitions were identified and rectified. Leveraging Monte-Carlo methods, enhanced system performance verification frameworks were established, culminating in increased confidence in the spacecraft's capability, especially during Earth fly-bys.

Data Engineering

Data analysis

Verification and Enhancement of the ESA Funded JUICE Spacecraft at Airbus Defense and Space

CMR Surgical

<p><strong>Customer Need:</strong> Medical organisations and clinical sites worldwide require accurate and timely insights from surgical telemetry data. The needs of this specific customer was no different. With the advent of advanced surgical robotics, the need for predictive maintenance and understanding of robot &amp; instrument health was paramount. Additionally, there was a demand for information regarding the fidelity of instruments, the health of robot joints, the overall co-bot experience, and the usage patterns across different clinical locations.</p><p></p><p><strong>Derived Solution:</strong></p><ol><li><p><strong>Automated Surgical Telemetry Analysis</strong>: Introduced algorithms that automatically analyze surgical telemetry data in the cloud.</p></li><li><p><strong>Predictive Maintenance with Deep Learning</strong>: Used deep learning techniques and extensive system knowledge to develop a predictive maintenance algorithm specifically for the robot wrist joint, enabling early detection of joint degradation from surgical data.</p></li><li><p><strong>Metrics Development</strong>: Acted as a core contributor in developing several key metrics derived autonomously from surgical data in the cloud. These metrics offered precise insights into various critical aspects of the surgical process and instrument health.</p></li></ol><p><strong>Value Delivery:</strong></p><ul><li><p><strong>Enhanced Instrument Longevity</strong>: With the predictive maintenance algorithm, clinical sites can now anticipate joint degradation in the V-Wrist joint, ensuring timely maintenance and potentially extending the lifespan of the instrument.</p></li><li><p><strong>Business-Critical Insights</strong>: The autonomously derived metrics from surgical data offered invaluable insights that fed directly into business-critical activities.<br><br>These insights aided in:</p><ul><li><p>Gauging instrument fidelity</p></li><li><p>Monitoring the health of robot joints</p></li><li><p>Understanding the co-bot experience from the customer's perspective</p></li><li><p>Analyzing usage patterns across clinical sites globally, facilitating better resource allocation and improved patient care.</p></li></ul></li></ul>

Leveraged deep learning methods combined with system knowledge to implement a predictive maintenance algorithm that can detect joint degradation wrist joint of a surgical robot, from surgical data.

ML Solutions for Predictive Maintenance and Analytics in MedTech

Slab Construction

<p>Le client utilisait Excel pour toute son activité: feuilles de temps, congés maladie, factures d'achats et de ventes, suivi des chantiers, des prestataires et leur documents légaux...</p><p></p><p>Cela devenait difficile à gérer, ne permettait pas l'accès concurrentiel par plusieurs employés, et l'erreur humaine étant facilement présente. Les exports comptables et financiers étaient également difficiles.</p><p></p><p>J'ai développé un CRM complet, dans lequl l'entreprise peut saisir les informations sur les chantierts, gérer leur suivi détat, y attacher des devis, attribuer des employés, des prestataires, des intérimaires...</p><p></p><p>L'ensemble des formulaires permet d'automatiser et de faciliter une grande partie de la saisie. L'outil permet d'afficher des statistiques financiers en plusieurs points du site, avec des comparaisons N / N-1, l'export automatique de la TVA (formatté, prêt à envoyer), l'export détaillé des heures travaillées, absnces, maladies, stages, le calcul automatiques des paniers et notes de frais, etc..</p><p></p><p>L'outil est en production depuis 1 an et des opérations de maintenance et d'ajout de fonctionnalités ont lieu de temps à autres.</p><p></p><div class="image-container"><div class="image-row-wrapper"><div class="image-wrapper"><img src="https://cdn.collective.work/cleyjy5mc0t7q5xhldc7ndnnf.jpg"><div class="image-overlay"></div></div></div></div><div class="image-container"><div class="image-row-wrapper"><div class="image-wrapper"><img src="https://cdn.collective.work/cleyjyh1e0tzg5wergh6v3qvn.jpg"><div class="image-overlay"></div></div></div></div><div class="image-container"><div class="image-row-wrapper"><div class="image-wrapper"><img src="https://cdn.collective.work/cleyjyr320t835xhl44i1077k.png"><div class="image-overlay"></div></div></div></div>

Cette société gérait toute sa comptabilité, ses feuilles de temps, factures d'achats et de vente, TVA, bilan comptable etc.. sur Excel. J'ai développé un outil plus fiable et plus performant, accessible en ligne et en accès concurrentiel pour les administrateurs de l'entreprise.

Javascript & Typescript

Laravel

Outil CRM complet pour une entreprise de construction à plusieurs centaines de millers d'€ de CA

Student Facilities

<p>The client needed a platform to help business schools student choose an abroad school, for the mandatory 1 year transition scholarship.</p><p>A lot of schools exist, but student had no objective data to compare them, except what the school wrote, which was most of the time about the location, price and program.</p><p></p><p>The student wanted to allow schools to say more about them, their campuses, what made them unique. I built several forms (around 50 questions) to help school manager bring the most of their schools, allowed students to leave review about the schools (reviews were moderated, and schools were able to respond, bringing the discussion to a new level), we also leveraged a job board, talked about everyday life, transport means, schools associations and extra scholar life...</p><p></p><p>The website is built on Laravel, HTML/CSS3/Jquery, multilingual, and hosted on Google Cloud Platform, with all security and backup automation needed.</p><p>That was a year long, 100k budget project, which i drove by myself</p><p></p><div class="image-container"><div class="image-row-wrapper"><div class="image-wrapper"><img src="https://cdn.collective.work/cleykak4d0u6f5wer2jdt7qh4.jpg"><div class="image-overlay"></div></div></div></div><div class="image-container"><div class="image-row-wrapper"><div class="image-wrapper"><img src="https://cdn.collective.work/cleykavxe0u6l5wer5oe5c8yi.png"><div class="image-overlay"></div></div></div></div><div class="image-container"><div class="image-row-wrapper"><div class="image-wrapper"><img src="https://cdn.collective.work/cleykb9ik0tfk5xhl4kim4lej.png"><div class="image-overlay"></div></div></div></div><div class="image-container"><div class="image-row-wrapper"><div class="image-wrapper"><img src="https://cdn.collective.work/cleykbfr20tfo5xhlfz7z4cz6.png"><div class="image-overlay"></div></div></div></div>

Business Schools student must do a 1 year transition scholarship abroad. Many business schools exist everywhere in the world, but students had no concrete and objective way to compare and choose them

Laravel platform to help international business school students choose their abroad university

DataSZ

<p>Several csv files were to aggregate (orders, products, customers, ...). Used pandas and feature engineering to analyze data and create relevant and effective unsupervised ML algorithm (K-means, DBSCAN, Hierarchical Clustering)</p><p>Used clusters and data to identify appropriate marketing opportunities</p>

Using Python and unsupervised algorithms to detect and create effective clusters and leverage marketing actions

Data vizualization

Strategy Consulting

Marketing clustering for a marketplace

<p>Developed Deep Learning models to predict when a robot could fail. Analyzed the business need, trained a Deep Learning model, tested it, and deployed on AWS. Result: the company saved 80% of its maintenance costs since launch of the solution.</p>

Machine learning

Predictive maintenance in robotics

<p>We are a multi-disciplinary team specialized in full-stack data science, from analyzing the business problem to training, testing, deploying, monitoring and continuously improving data-based models. We thrive to work with the most rigorous scientific approaches to deliver high quality solutions.</p><p>Always willing to share our knowledge, we organize conferences about our work.</p>

AI Engineer

Multidisciplinary Software Engineer

Senior Data Scientist

We develop Deep Learning applications for computer vision with real-time performance, cloud deployement and integration.

Real-time Computer Vision

Modern and effective ETL pipelines and data stacks on Azure, AWS and GCP. Experts at delivering MLOPS solutions, from training to monitoring and contiunous improvement and integrations of your models

We analyse data with the most advanced tools and rigourous scientific approaches  in order to solve your business or engineering problems and bring value with predictive models.

Data Science

We help you on your path to digital transformation in order to scale your business with the right data infrasctruture and workflow. We also teach your teams on how to use the most efficient data tools

757 collectifs