Youssef Mahfoud

Within CATIA R&D - 3D for Collaborative environments - Designed and implemented synthetic dataset generation framework using CATIA's internal C++ APIs and unit tests. - Generated over 10k images of 8 distinct 3D models each with multiple parts and from different views, all while generating a JSON file along to help the training later. - Developed, tested, analysed and iterated deep learning architectures ( +10 itterations in architechture using CNNs/ViTs ) and training pipelines ( Pretraining + Finetuning/TransferLearning ) using PyTorch and CUDA for accurate DSPBR material encoding, accounting for lighting conditions and 3D geometry using multiview images. - Using the trained encoder as a solid ( image+mask -> vector ) material embeder, in order to perform vector search with end-user catalog materials. - Integrated the solution into a WebApp enabling material transfer from ControlNet-generated renderings to end-user catalog materials applied to each part of any 3D model in 3DExperience's CATIA.

Built image and video generation pipelines using ComfyUI and Automatic1111. Integrated custom models, LoRAs, IPAdapters and ControlNet for brand-specific and precise visuals. Facilitated the integration of AI-generated assets into graphic designers' workflows, improving collaboration and efficiency. Stayed up-to-date with research papers and theoretical advancements to guide implementation.

NeRF-Based Multiview Mammography for Breast Cancer Diagnosis: (3D Detection using multi-view 2D GTs) Used multi-angle scan images to reconstruct the 3D breast. Developed a 3D anchorbased RPN (Region Proposal Network) optimized using 2D ground truth boxes. Designed an itterative projection algorithm to map 3D boxes to 2D for loss optimization. Constructed a dynamic loss function that adapts to the number of NeRF training images with 2D ground truths, making it generalizable to various imaging types and scenes.

Seating department

Skills: Adobe Photoshop · Adobe Illustrator · Adobe Premiere Pro · Behance · After Effects · Visual Effects · Video Editing