At INRIA Sophia Antipolis, I built a 300-processor Linux HPC cluster and supported researchers with parallelization and code optimization. I also managed the GRID5000 technical platform for parallel computing.
As a freelance software engineer, I optimized computer vision and deep learning pipelines, including a Snap Research OpenCV/PyTorch pipeline with CUDA and TensorRT. My client work has also included CUDA-based RAW-to-EXR conversion for Lightcraft Technology and satellite image processing for 3D reconstruction at Airbus.
At REALViZ, I co-founded a visual effects software company and worked as software architect, contributing to MatchMover and a digital image processing API. I also teach CUDA, and my work spans GPU computing, real-time 3D graphics and image processing.

