Ons Ouhibi

- Conducted a comparative analysis of web scraping tools to extract and structure legal content from diverse online sources. - Designed and implemented advanced Retrieval-Augmented Generation (RAG) pipelines, including Naive, Corrective, Fusion, and Agentic RAG, to enhance the contextual accuracy and relevance of generated responses. - Developed a chatbot with PDF Q&A, web search, and deep search features. - Evaluated RAG system performance using the RAGAS framework, focusing on key metrics such as faithfulness, relevance, correctness, and semantic similarity. - Integrated large language models (OpenAI, Hugging Face) into LangChain-based workflows to build intelligent retrieval and question-answering systems.

Developed an AI solution for information extraction from CVs using three distinct approaches: – Named Entity Recognition (NER) using SpaCy: Employed NER and NLP techniques to identify and extract relevant information from CVs. – Large Language Models (LLMs): Leveraged Gemini API to enhance extraction precision and accuracy. – Local LLM Integration: Successfully integrated and optimized local LLMs to further refine the extraction process.

-Developed three advanced models for automated document classification: -National ID Card detection model using the "One-Class SVM" approach. -PDF classification model employing "Logistic Regression" techniques. -Image classification model based on Convolutional Neural Networks (CNN) for enhanced accuracy. -Utilized Python, Scikit-learn, TensorFlow, and OpenCV to implement and optimize these models. -Contributed to enhancing document processing efficiency through innovative machine learning solutions.