AI Projects

Motion Capture Data Cleanup

An AI-powered system for automated cleanup of motion capture data using deep learning techniques. Researched papers and implemented LSTM based solutions. Designed new .c3d data ingestion pipeline and retraining.

Tech Stack: Python, PyTorch, PyTorch Lightning, Vicon Shogun API

3D Pose Detection and Object Tracking System from Monocular RGB

Master's thesis project: A real-time 3D object detection and tracking system for WebAR applications, featuring real-time rendering of 3D mesh on top of tracked objects using monocular RGB input.

Tech Stack: Google Mediapipe, WebGL, Three.js

Hawkeye: Neural Image Search Engine

A content-based image search engine developed for the Multimedia Information Retrieval and Computer Vision course. Uses query-by-example paradigm with advanced neural feature extraction and efficient spatial indexing.

Tech Stack: Python, TensorFlow, MobileNetV2, Flask
Key Features:
  • • Vantage Point Tree for efficient similarity search
  • • Dual feature extraction using base and fine-tuned MobileNetV2
  • • Interactive web interface for image queries
  • • Comprehensive performance evaluation system

Medical Image Analysis CNN

Developed a Convolutional Neural Network for detecting and classifying abnormalities in mammography images. Part of the Computational Intelligence and Deep Learning course at University of Pisa.

Used the Modified CBIS DDSM (Curated Breast Imaging Subset of Digital Database for Screening Mammography) dataset for training and validation.

Tech Stack: Python, PyTorch, OpenCV, scikit-learn

GIAR: Automatic Game Genre Prediction

An advanced feature for the GIAR platform that uses machine learning to automate the tagging of game genres based on game descriptions. Developed for the Advanced Data Mining and Machine Learning course, this extension speeds up the game data entry process, offering genre predictions that admins can review and adjust.

Tech Stack: Python, Scikit-Learn, FastAPI, Rawg API
Features:
  • • Automated genre prediction for new game entries
  • • Multilabel classification based on game descriptions
  • • Integration with Rawg API for comprehensive data
  • • Customizable genre suggestions for admins
  • • Simplifies and accelerates game data management

DBMS Projects

GIAR: Multi-Database Gaming Platform

A comprehensive gaming platform developed for the Large-Scale and Multi-Structured Databases course, featuring both document-based and graph-based data storage solutions. Later extended with machine learning capabilities for the Advanced Data Mining course.

Tech Stack: MongoDB, Neo4j, Python, FastAPI
Features:
  • • Hybrid database architecture (MongoDB + Neo4j)
  • • Document-based game data storage
  • • Graph-based user relationships and recommendations
  • • Comprehensive index performance analysis
  • • RESTful API service integration

Cloud Projects

Distributed K-Means Clustering

Implemented the K-Means clustering algorithm using both Hadoop MapReduce and Apache Spark frameworks as part of the Cloud Computing course at University of Pisa. The system efficiently processes large datasets by parallelizing the clustering computations.

Tech Stack: Hadoop, Apache Spark, Java, Python (validation)
Features:
  • • MapReduce workflow implementation
  • • Custom combiner for optimization
  • • Parallel processing of clustering iterations
  • • Validation using scikit-learn generated datasets

Distributed Memory Monitoring System

A cloud-based monitoring system developed for the Cloud Computing course that tracks memory usage across remote machines. The system implements a producer-consumer architecture to collect, store, and visualize memory metrics using Gnocchi time-series database.

Tech Stack: Python, Gnocchi, SSH, OpenStack
Features:
  • • Producer component for remote data collection via SSH
  • • Consumer component for data visualization
  • • Time-series data storage using Gnocchi
  • • Periodic memory usage monitoring
  • • Distributed system architecture

VR Projects WIP

VR Museum - OpenAI NPC - Unity-OpenAI comunication - Server

Worked in the OpenAI integration in Unity for a realtime NPC-user comunication in Unity.

Tech Stack: Unity, C#, OpenAI API, FastAPI, Docker, OpenStack, Python, WebSocket

Cybersecurity Projects

Secure Multi-Client Messaging System

A secure client-server application developed for the Foundations of Cybersecurity course. Features encrypted communication channels, digital signatures, and multi-threaded architecture for handling concurrent client connections.

Tech Stack: C++, OpenSSL, POSIX Threads
Key Features:
  • • Message encryption and digital signatures using OpenSSL
  • • Multi-threaded server for concurrent client handling
  • • Secure authentication protocol
  • • Real-time message forwarding between clients

Extra Projects

Cosmic Particle Detector

Educational tool built during University of Hawaii internship for detecting cosmic particles using Raspberry Pi. Integrated sensors and components: Hacked webcam, GPS sensor, LEDs, image analysis algorithms

Tech Stack: Python, Raspberry Pi, OpenCV

Home Assistant Integration

Bachelor's thesis project: Implementation of the LWM2M standard in the Home Assistant Hub for IoT device management.

Tech Stack: Python, Home Assistant API