Leila Skouri

• Implemented Hidden Markov Models (HMM) to discover latent patterns and state transitions in CRM data, enabling advanced customer behavior analysis. • Developed XGBoost supervised learning pipeline for multi-class classification of customer next actions (stay, forward, backward progression). • Engineered 3 bidirectional ETL pipelines using SSIS: web scraping extraction, flat file to SQL Server (SSMS), and SSMS to Dynamics 365 CRM integration for clean and structured data flows. • Integrated LangChain chatbot using Llama 3 for natural language querying of project data and interactive insights. • Created Power BI dashboards with multiple KPIs for real-time monitoring and business decision-making. • Built .NET Core web application with Microsoft Cookie authentication and role-based access control (RBAC), delivering a full-stack data-driven solution.

Engineering Internship at Tenstep via Esprit Engineering School Developed an AI-powered Content-Based Recommender System for Project Risk Management to help project managers identify, assess, and mitigate risks effectively. Key Implementations: • Built a content-based recommender system using Natural Language Processing (NLP) to extract and structure Project Risk Management data from various sources. • Developed a conceptual knowledge graph using Neo4j for personalized, PMBOK-aligned risk recommendations tailored to specific project contexts and methodologies. • Created an interactive React/Flask web application providing real-time risk mitigation insights and decision support through an intuitive user interface. • Implemented intelligent recommendation algorithms that analyze project characteristics and suggest relevant risk management strategies based on industry best practices.

Explored and implemented advanced unsupervised clustering algorithms Affinity Propagation, CURE (Clustering Using Representatives), and OPTICS (Ordering Points To Identify the Clustering Structure) for customer segmentation using data from Dynamics 365 Sales. The project aims to classify customers into meaningful groups to inform targeted marketing strategies, personalized customer engagement, and improved sales processes.

Integrated Project via Esprit Engineering School for PGS Developed an AI-powered Forecasting System for Inventory Management as part of an academic-industry collaboration project between Esprit and PGS. Key Implementations: • Built order prediction and customer segmentation models using Artificial Neural Networks (ANNs) to analyze purchasing patterns and customer behavior. • Implemented demand quantity forecasting using FeedForward Neural Networks combined with YOLO object detection for real-time inventory tracking and prediction. • Optimized gas/oil distribution queues by applying Dijkstra's shortest path algorithm to minimize wait times and improve logistics efficiency. • Followed Team Data Science Process (TDSP) methodology for end-to-end ML pipeline development, ensuring structured and reproducible workflows.

Implemented interactive dashboards in Power BI to monitor KPIs and optimize business decision-making processes