GUETTI Mouhsine

Built a RAG system: scraping artticles and paper from the intenet, text cleaining and pre-processing(classification, Chunking, Embedding and vectorization..), powered by LLM for prompt answering. Implemented and fine-tuned LLMs (Mistral, LLAMA2, Qwen) for GPU & CPU inference through QLoRA quantization, PEFT fine-tune, and LLAMA.CPP optimization. Built an OCR system to extract and label document data using NER and LLM techniques. Designed and deployed a web application interface for seamless API interaction.

Applied digital image processing techniques, including image cleaning and enhancement. Designed and implemented deep learning models, such as ResNet, GANs, and diffusion models, for advanced analysis. Explored the use of SuperPoint and SuperGlue networks for feature detection and matching in hyperspectral data. Ensured research quality through rigorous methodology, reliable data collection, and robust analysis. Tools: PyTorch, Transformers, Spark, OpenCV, scikit-image, XGBoost, ArcGIS.