Angelo Guarino

Driven by curiosity for understanding and advancing intelligent systems.

Research interests: Natural Language Processing • Large Language Models • Model Reasoning • Multi-Agent Systems • Explainable AI • Computer Vision

Research

Graduate Researcher | NLP Group, University of Padova | March 2025 - Present
Supervised by Prof. Giovanni Da San Martino

Building on the framework of SemEval 2025 Task 10, I am designing NLP pipelines that leverage LLM-based extraction and hierarchical clustering to automatically build multi-level narrative taxonomies from large news corpora.

My work includes developing "narrative-aware" summarization techniques that capture and condense persuasive content using hierarchical narrative structures.

Publication in preparation: Guarino, A., Sartori, E., & Da San Martino, G. "Hierarchical Narrative Understanding via LLM-Guided Clustering" (targeting ACL Rolling Review).

Work

Previously, I worked at Cubbit as an SDR and a Data Scientist.

Technical Skills

• Model Development: Designed novel architectures (FineGAN for inpainting), fine-tuned LLMs (Phi, Llama, Gemma) using PEFT methods, built custom models in PyTorch for Kaggle competitions
• NLP & Language Understanding: Hierarchical topic modeling, LLM-guided clustering, multilingual text analysis, narrative-aware summarization
• Theoretical Foundations: Convex/non-convex optimization, stochastic methods, statistical learning, information theory
• Systems & Engineering: Python, PyTorch, LangChain, Weights & Biases, Gradio, Numpy, Pandas, Git, LaTeX, Docker