Mehdi Oudghiri

- Developed representation learning approaches for document classification using embeddings derived from neural networks. - Integrated OCR, layout analysis, and NLP techniques to extract and structure information from heterogeneous geotechnical reports. - Conducted experimental evaluations to compare unsupervised models and feature extraction strategies.

- Conducting research on CAPTCHA robustness using object detection models (YOLO) in a security context. - Evaluating the effectiveness of image-based challenges through computer vision and machine learning approaches. - Analyzing system vulnerabilities and limitations in real-world adversarial settings.