Nourchene ROMDHANI

– Analyzed multi-sensor data to evaluate repeatability and stability, enabling automated sensor state diagnostics. – Studied the impact of water exposure on sensor readings and characterized the resulting response and drift patterns.

– Developed AI systems for a BMW automotive seat system, including Occupant Classification (OCS) and Weight Estimation using multi-sensor seat data. – Built classification and regression models using supervised learning, with feature engineering and data preprocessing for robustness in real-world conditions.

– Designed and developed an AI agent customization and enhancement platform for automotive seat systems using Large Language Models (LLMs), Model Context Protocol (MCP), and MQTT. – Implemented a YAML-based configuration system and Seat Agent Studio interface to enable AI agent customization, extension, and tool management. – Improved agent reasoning capabilities and tool performance through independent enhancements. – Implemented a local deployment architecture allowing the AI agent to run independently from the platform, improving performance, latency, and system autonomy.

– Built a real-time Big Data pipeline using Kafka and Apache Spark to stream and analyze NASDAQ market data from Yahoo Finance APIs. – Leveraged HDFS and Cassandra for distributed storage and developed Streamlit dashboards for live financial insights.

– Developed an intelligent accounting assistant using LLM and RAG to analyze financial documents, extract key data, and detect anomalies. – Integrated semantic search and a conversational interface for automated and efficient accounting management.

– The Eye-Controlled Virtual Keyboard project aims to develop an assistive typing system controlled by eye gaze and blinking. – The system tracks eye movements in real-time to navigate the keyboard interface, while selections are confirmed by blinks.

– Developed an AI-based system to predict customer churn and recommend similar high-risk clients for targeted retention strategies. – Implemented XGBoost with GridSearchCV for predictive modeling and a content-based recommendation engine using TF-IDF and cosine similarity.

– Healthcare portal mobile app with appointment management, secure communication, and AI-based medical report analysis.

– The GuideMe project aims to develop a smart mobile application for tourists in Tunisia. – Using object detection and augmented reality, the application displays an interactive avatar that provides historical and cultural information about detected monuments and objects.

– Customer support chatbot providing automated assistance, inquiry handling, and request management.