Interactive visualization framework for explainable AI that coordinates spatial neighborhood projections with hierarchical decision tree visualizations, enabling intuitive exploration and validation of local rule-based explanations through bidirectional interaction mechanisms.
Hackathon project exploring Explainable AI techniques (LIME, SHAP, LORE) to understand black-box models in marketing uplift modeling, evaluating explanations across content, presentation, and user experience dimensions for diverse stakeholders.
Proposed enhancements for Fondo Ambiente Italiano website's user experience through user research, need validation, and solution development. Designed and tested three innovative solutions using structured assessment methodology to improve event discovery.
Data warehouse and OLAP solution analyzing Chicago traffic crash data to support mock insurance company decision-making through multidimensional analysis, ETL pipelines, MDX queries, and interactive Power BI dashboards.
Expansion of data mining I project analyzing Spotify music data through time series classification, neural networks, and ensemble methods. Applied DTW, Random Forest, and Gradient Boosting for prediction tasks, with LIME, LORE, and SHAP explainability revealing audio features relationships.
Strategic intelligence analysis of coding trends through social media, analyzing 3.8M+ tweets, Dev.to articles, Reddit posts, and custom-built R scraper for extracting structured Q&A data to identify technology adoption patterns and emerging methodologies.
Forecasting system for banking sector EBIT and Net Income predictions using machine learning models trained on financial statements from Italy's top 5 banks (2018-2022) and macroeconomic indicators.
Business process modeling project analyzing a painting school scenario using BPMN and Workflow Nets. Formal verification of process properties including soundness, boundedness, and liveness using Petri net analysis and WoPeD tool.
Implementation and comparison of Lagrangian dual-based optimization algorithms using Nesterov's Fast Gradient Method with dynamic smoothing for solving quadratic minimum cost flow problems on network graphs.
Recreation of a paper on statistical analysis of word embeddings that found connections between Word2Vec and PMI matrices.
Data mining project analyzing 15,000 Spotify tracks using clustering, classification, regression, and pattern mining techniques to predict songs metadata.
A comprehensive machine learning and XAI pipeline for predicting TV show ratings using tabular and textual data, utilizing CatBoost and SHAP/LIME for model interpretability.
Full-stack Italian-language dog shelter adoption platform built for Canile Sanitario Bari. React 19 + TypeScript SPA on Cloudflare Workers (Hono backend, KV data store, R2 media), with an admin dashboard, WhatsApp-first contact flow, Instagram caption generator, and live analytics — all on Cloudflare's free tier.
Civic data analysis platform for Bari, Italy, processing 1,000+ survey responses through AI-powered NLP, geographic heatmaps, and statistical correlation analysis to transform citizen feedback into actionable insights.
Interactive web map platform for visualizing aggregation spaces, social struggle locations, and study environments across my home town (Bari). Developed for an association to document youth infrastructure and support community access to urban spaces.
Interactive web-based visualization toolkit for urban research in my home town (Bari). Developed a D3.js survey dashboard with multiple chart types and a Leaflet.js mapping tools.
Machine learning system for detecting student moods (engaged, confused, frustrated, bored, drowsy, looking away) in e-learning environments through facial Action Unit analysis.