Abdelaziz Ouayazza

Developed a comprehensive framework for automatic peak detection in two-dimensional NMR spectra. Designed and implemented an end-to-end NMR data processing pipeline, from Bruker FID files to automated peak detection using TopSpin, NMRGlue, and Python. Developed a realistic synthetic HSQC spectrum generator with automatic ground-truth annotation generation for supervised learning. Designed, trained, and optimized a U-Net semantic segmentation model (U-NetPicker) for automated peak picking in 2D NMR spectra. Developed and benchmarked several classical feature detection methods, including Harris, SUSAN, LoG, DoG, Hessian, Local Maxima, and the Electrostatic Model, combined with Watershed segmentation and DBSCAN clustering techniques. Evaluated the performance of the proposed methods on both synthetic and experimental datasets (HMDB) using Dice Score, Precision, Recall, and F1-score metrics. Conducted an extensive literature review on Artificial Intelligence for NMR spectroscopy, automated peak-picking methods, and multidimensional NMR spectrum analysis