Responsible for the development of a medical image segmentation project using deep learning techniques at Visible Patient, aimed at detecting anatomical structures within medical images (DICOM). My role encompasses the training, inference, and evaluation phases of algorithm development.The project is organized into several sub-projects:- VP Deep Learning: Training of AI models (based on U-Net architecture), 3D images preprocessing and metrics for result quality evaluation.- Optimized Inference Server: Development of a server accessible via a REST API, along with workers for segmentations and post-processing execution. Deployment is managed both on-premise and on AWS.- Clinical Evaluation: Involvement in the clinical evaluation of segmentation results generated by AI (metrics production, result interpretation).Development adheres to medical device standards (IEC 62304, ISO 13485), with rigorous risk management, testing, and traceability protocols.Technical Environment: Python, NumPy, CuPy, SciPy, Pandas, Keras, TensorFlow, Celery, MinIO / S3, GitLab CI, AWSTeam of 3 to 4 members

Project Lead for a Medical Image Segmentation Scheduler, designed to optimize the processing of these images through an AI developed by Visible Patient for the production service.- Collaborative Design: Created prototypes and detailed specifications in close collaboration with end users.- Project Management: Drafted specifications, planned and coordinated project milestones, managed user feedback, and addressed incident handling.- Application Development: Contributed to development (backend, frontend) and interfacing with other services within the ecosystem.Technical Environment: TypeScript, Express, Angular, Docker, OpenID Connect, MongoDB, MinIO, GitLab CITeam: 2 people

Development of the VP Planning Mobile for Android app, designed to provide practitioners with advanced 3D visualizations of Visible Patient models, thereby facilitating surgical planning.Key Responsibilities:- Designed and developed the application (Kotlin)- Integrated complex 3D visualizations (VTK)- Automated tests (unit and integration tests)- Setup of the CI on GitLab to automate builds and tests- Designed and implemented reactive architectures, ensuring the app’s responsiveness and performance- Development adhered to medical device standards (IEC 62304, ISO 13485), with rigorous risk management, testing, and traceability protocolsTechnical Environment: Android, Kotlin, RxJava/RxKotlin, Realm, Retrofit, VTK, Android Studio, Docker, GitLab CI

https://play.google.com/store/apps/details?id=com.sweechDesign, development and publishing of a file sharing app for Android: SweechServer part (Android) :- JSON API based on Servlets exposing the content of the mobile- Security aspects (HTTPS, IP filtering, ...), Advanced Android integration (Power management, Sharing)- URL sharing through Android Beam (NFC) / QR CodeWeb interface :- UI based on WebComponents (Polymer)- Advanced File transferPromotion:- Publication on Play Store- Presentation videoTechnologies: Android, Java, BouncyCastle, Javascript, HTML5, Polymer, WebComponents, HTTP, Servlets, NFC, QR Code, Blender

Four J's develops a toolkit and an infrastructure tailored for business applications development. A virtual machine provides an abstract representation of the interface to display (DOM/CSS like). This is interpreted by several clients (Web, Desktop, Android or iOS) to interact with users.Genero Web Client- Architecture and development of an in-browser Genero rendering engine- Toolkit API design- Continuous integration setup (Protractor - Jasmine)- Collaboration with our support centersTechnologies : Javascript, HTML5, SASS, CSS, nodejs, grunt, NPM, protractor, selenium, git, JIRAGenero Mobile for Android (http://www.generomobile.com)- Architecture et developement of the product- Continuous integration setup- Scrum master- Collaboration with our support centersTechnologies: Android, Java, C, NDK, gradle, multithreading, HTTP, git, JIRAGenero Application Server- Redesign and evolution of the web rendering engine- Maintenance and evolution of our in-house templating language interpreter (DSL)- Optimisation and performance - Collaboration with our support centersTechnologies : C++, boost, socket, TCP, multithreading, IPC, POSIX, Win32, Apache APR, HTML5, Javascript, svn, bugzillaGenero Studio- Developement of the main framework on which the IDE for the Genero language is based.Technologies : C++, Qt, Design Patterns, svn bugzilla

Member of the BH Team (http://www.bhteam.org) involved in the French Robotics Cup.• Development and maintenance of a custom embedded linux based OS specially tuned for our hardware platform (hacked Seagate Dockstars). Full system with toolchain below 100Mo. (Linux kernel patches, buildroot, cross-compilation)• Development of computer vision modules (C++, OpenCV)• Developement of an event based execution framework (Brewery) in Python based on asyncore. The objective was to be able to write asynchronous code linearly (coroutines)• Developement of a robot simulator (Python, PyQt)https://bitbucket.org/bhteam/bhware-openResults on over 160 teams:• 2015 - 9th• 2014 - Quarter finalist• 2013 - 19th• 2012 - 9th• 2011 - 2th• 2010 - 59thTechnologies: linux kernel, buildroot, C, C++, OpenCV, Python, PyQt, simulation, embedded, microcontrollers, coroutines

France Télécom R&D (Today Orange Labs) : Development of 2G and 3G networks radio planning and optimisation tools- Interfacing of a radio frequencies allocation and optimisation engine with a radio engineering platform (Planet Evolution) (C++ / C# / .NET).- Development the UI for the optimisation engine (C++ / Qt).- International environment (collaboration with a team based in Canada)

- Study and development of optimisation algorithms for 3G networks radio frequencies (C++, Python).

- Mise en place d’un SGBDR Interbase- Conception d’une base de données et migration de diverses bases existantes vers ce nouveau système- Réalisation d’un programme d’exploitation statistique et de reporting- Développements en C++ et SQL sous Borland C++ Builder