Alessandra Lumini

Design a RAG system for secure management of confidential documents
Development of an AI system to manage confidential documents, with authenticated access, granular permissions, and audit trail, aimed at reducing administrative workload and ensuring compliance with privacy regulations.

Design and test local LLM solutions for coding
Experimentation and validation of large language models (LLMs) executed locally, focusing on code generation and programming assistance. Includes installation, configuration, performance testing, and integration in controlled development environments.

Design a mobile application for indoor user tracking with integration to the notification system
Development of a mobile app to track user positions within indoor environments using indoor localization technologies. The app must communicate with the existing notification system, ensuring reliable operation even under degraded or absent connectivity, and collect movement data for managing dynamic scenarios.

Computer vision in sports: ball tracking in beach volleyball.
Development of a computer vision application for ball tracking in beach volleyball videos. The ball is a small, fast-moving object, often subject to motion blur, occlusions (hands, net, players), and scale changes (zoom/framing). Reliable tracking is a prerequisite for advanced analysis: event detection (serve, set, spike), statistics, rally segmentation, and technical KPIs.

Computer vision in sports: player tracking in beach volleyball.
Development of a Multi-Object Tracking (MOT) system for continuous monitoring of players on the sand. The thesis addresses challenges related to extreme poses during dives, teammate re-identification, and managing body intersections (overlap). The goal is to calculate athletic metrics such as movement speed, defensive space coverage, and coordination between team members during offensive and defensive phases.

Development of an interactive web platform for performance analysis in beach volleyball.
Design and implementation of a web dashboard to monitor and analyze sports performance, collect granular data from players and matches, and generate advanced statistics and dynamic visualizations (heatmaps, performance charts). The system provides decision-making tools for strategy improvement, allowing coaches to filter key match moments and compare historical athlete data.

Design multidimensional data models for dashboards and dynamic reports
Development of data models aimed at visual analysis and dynamic report production, using advanced Business Analytics tools on a reserved cloud platform.

Develop Java components for microservice-oriented ETL architecture
Design of Java libraries and components to capture, filter, and process data from NoSQL sources (MongoDB) in a microservices architecture, with attention to modularity and scalability.

Design predictive data models based on administrative data
Development of statistical models to identify patterns and fluctuations in the entity’s procurement and tender data, aiming to support strategic decisions and assess impacts on the budget.

Design a data lake architecture for managing heterogeneous data sources
Definition of technical solutions to centralize heterogeneous data sources into a single access pattern, evaluating challenges and potential from the perspective of distributed and scalable ETL/ELT infrastructures.