Zaineb Boujelbene

– Designed and deployed an enterprise Multimodal SQL AI Agent using RAG architecture and LLMs, enabling automated natural-language-to-SQL query generation for business users. – Built a conversational AI system handling French/English queries using Ollama LLM and RLHF feedback loops, improving query accuracy and response relevance. – Implemented schema-aware Retrieval-Augmented Generation (RAG) leveraging database metadata, enhancing contextual SQL generation and reducing query errors. – Developed multimodal document analysis (Excel, PDF, Word) using vision models for automated content extraction and structured data processing. – Integrated n8n automation workflows for scheduled database archiving, improving operational efficiency and reducing manual intervention. – Delivered full-stack deployment using FastAPI, Angular, WebSockets, and SQL Server with CI/CD via GitLab Actions.

– Developed a Generative AI-powered smart agricultural assistant using LLMs and IoT data streams for crop health monitoring and decision support. – Built predictive ML models (TensorFlow/Keras, Scikit-Learn) for pest/disease detection and resource optimization, enabling data-driven farming decisions. – Integrated IoT sensors (ESP32, MQTT) with cloud infrastructure for real-time data ingestion and model deployment. – Designed a React-based web interface with LLM-powered chat assistant to provide real-time recommendations and alerts to users. – Automated CI/CD workflows using GitHub Actions to streamline deployment and model updates.

– Developed a using OpenCV and TensorFlow for automated employee entry/exit tracking. – Integrated facial detection and recognition models with a database to generate timestamped attendance logs. – Built a Flask + React web application to manage and visualize attendance records, reducing manual HR tracking processes. – Implemented OCR capabilities to enhance data extraction and system reliability.

– Developed a Customer Churn Prediction model using data transformation, feature selection, and hyperparameter tuning techniques. – Applied deep learning models (Keras, TensorFlow) to improve classification performance and customer retention insights. – Contributed to research publication on churn prediction enhancement methods.

– Built web scraping pipelines using Selenium and BeautifulSoup to collect structured call center data.

– Developed a forest fire prediction tool using machine learning models based on the Canadian Fire Weather Index approach.

– Applied Scikit-Learn and TensorFlow for predictive modeling and risk classification. – Implemented geospatial visualization with Cartopy to support map-based decision-making for fire risk monitoring.