R&D Software Engineer

Designed and developed artificial intelligence applications, a distributed pre-segmentation environment, and online services for medical image analysis and modelling.

Built a distributed server for automated medical-image pre-segmentation, including REST APIs, asynchronous task queues, and storage for images and annotations. Developed a supervised deep neural network tool suite for training, validation, monitoring, experiment versioning, and multi-format model export.

Worked on medical image synthesis through deformable registration using semi-supervised deep neural networks. Optimized image transforms and model inference for GPUs using CuPy and TensorRT, including FP16 and INT8 quantized inference.

Developed software in compliance with IEC 62304 and ISO 14971, producing technical and regulatory documentation, documented architecture, requirements traceability, SOUP management, and verification materials. Also created internal tools for document anomaly detection, quality-of-working-life survey management, and automated interface testing.

Developed automatic urban-image segmentation solutions using deep neural networks within the Inno'lab innovation laboratory.

Adapted and trained convolutional semantic-segmentation architectures inspired by U-Net, V-Net, and SegNet. Worked with a public autonomous-driving dataset containing 25,000 annotated urban images across 66 object categories.

Conducted a literature review of existing methods and compared experimental results with reference models.

Contributed to the open-source Go API for the CCSDS Mission Operations standard, focused on the Message Abstraction Layer.

Developed archive-service storage for COM objects defined by CCSDS recommendations. Produced technical documentation and a demonstration program showing the service in a real-world use case.

Published the open-source code, which was subsequently incorporated into an official CNES repository.