Senior Research Engineer and AI expert specializing in optimization and deployment of high-performance Deep Learning models. PhD in Engineering from Vrije Universiteit Brussel (VUB), awarded with Greatest Distinction for research in deep learning optimization. Professional focus centers on bridging theoretical research with industrial-scale systems. Founder of a 3D Gen-AI startup BeViAI and former Senior ML Engineer at PicsArt.
Introduced a novel compression framework to reduce communication overhead in distributed training. This work enables faster scaling of large-scale models across clusters by optimizing how gradients are exchanged between nodes.
Developed a specialized loss function designed to improve the performance of lightweight neural networks on edge devices. This research directly addresses the accuracy-latency tradeoff for mobile-first AI applications.
An open-source implementation of a multimodal pipeline for generating temporally consistent human avatars. The system leverages audio and visual inputs to synchronize lip movements and facial expressions for high-fidelity digital humans.