Nizar EL MOKHTARI

- Development and optimization of a classification model based on the ResNet50 architecture to identify high-risk wind turbines, achieving a fine-tuned 99.7% accuracy on the F-beta score metric using semi-supervised techiques. - Data processing and ingestion with PySpark, development of advanced dashboards using PowerBI - Creation of processing pipelines with Dataiku and Palantir, integration with AWS - Generation of automated reports using PowerBI Builder

Developed a GenAI application to assist sales managers in analyzing prospects and generating pitches based on a long-term memory language model (LLM) to provide recommendations and relevant suggestions for each stage of the sales process. • Identified the target client and potential prospects, collected data via web scraping, and prepared CSV files containing contacts and prospect information. • Transformed the data into embeddings and stored them in a Vector DataBase, and used RAG to integrate the data into LLM via LangChain. • Generated sales pitches and created dynamic graphics from the data analyzed by the ChatBot GenAI. These elements were then integrated and generated into PowerPoint presentations.

During my internship at ToumAI, I actively participated in the development of a platform dedicated to analyzing African data and detecting weak signals within these vast datasets. Our approach included the following steps: - Collection of African data from social networks (using Selenium). - Implementation of a language identification model to handle linguistic diversity (using SVM and XGBoost). - Application of topic modeling techniques and clustering of themes (using BERTopic and LDA). - Detection of weak signals in the data.