Evidence
Projects
Each project documents the problem, data, methodology, results, and limitations. Real numbers only, no invented metrics.
Deep Learning · Time Series
Temporal Characterization and Deep Learning-Based Modeling of Human Sleep
Modelled temporal sleep physiology from polysomnography signals and compared deep learning against gradient-boosted sequence models for sleep-stage prediction.
Bi-LSTM reached 84% accuracy (F1 0.65) vs. 74% (F1 0.57) for the XGBoost baseline.
View projectComputer Vision · Deep Learning
MARIDA: Marine Debris Detection from Sentinel-2 Satellite Imagery
Built a PyTorch multi-label classification model to detect marine debris and ocean-surface features across 15 imbalanced environmental classes from satellite imagery.
Strongest performance on Marine Water detection (AP 0.63); macro F1 0.19, micro F1 0.23, mAP 0.12 overall.
View projectNLP · RAG
Airline Review Retrieval-Augmented Generation System
Built a retrieval-augmented generation pipeline over airline customer reviews to surface actionable insight from unstructured feedback.
Deployed a working RAG pipeline with vector search, prompt engineering, and a Streamlit interface.
View projectApplied AI · Full-Stack
Home Services Intelligent Chatbot
Developed a full-stack chatbot for booking home services with intelligent triage, dynamic pricing, scheduling, and bilingual support.
Shipped a deployed, full-stack academic implementation covering triage through booking confirmation.
View projectStatistical Modelling · R
Impact of Land Use & Habitat Features on Bird Diversity
Applied regression modelling to assess how land use and habitat features affect bird species richness and functional diversity in Northern Italy.
Produced regression-based findings on land-use effects, visualized for a non-technical audience.
View project