Andrea Bacciu
Applied Scientist II @ Amazon · Ph.D. in NLP
I am an Applied Scientist II at Amazon in Madrid. I am the science lead for Cross-lingual Customer Review Summarization and I work on LLM-as-a-Judge, specifically researching Uncertainty Estimation.
I hold a Ph.D. in NLP from Sapienza University of Rome, supervised by Professor Fabrizio Silvestri. My work on Cross-lingual Semantic Role Labeling was recognized with the 🏆 NAACL 2021 Outstanding Long Paper Award.
My journey in NLP spans from classic TF-IDF, Word2Vec, BERT up to LLMs. Along the way, I have worked on a diverse range of tasks including Semantic Role Labeling, Semantic Parsing, Word Sense Disambiguation, Sequential Recommender Systems, and Graph Neural Networks.
Interests: Uncertainty Estimation · Information Retrieval · NLU
Experience
Amazon Stores — Applied Scientist II, Madrid 2023 – Present
- Science lead for Cross-lingual Expansion of Customer Review Summarization CX
- Research on Uncertainty Estimation for Large Language Models
Amazon Alexa AI (AGI) — Applied Scientist Intern, Cambridge 2022
- Initiated research on Uncertainty Estimation for LLMs
- Published work on ontology gap handling in semantic parsing (*SEM 2024)
Education
Ph.D. in Natural Language Processing 2021 – 2025
Sapienza University of Rome · Advisor: Prof. Fabrizio Silvestri
Thesis: “Beyond Traditional Search: Bridging Retrieval, Reasoning, and Language Barriers in Intelligent Search Systems”
Funded by the National AI Competition research grant
M.Sc. in Computer Science (with honors) Sapienza University of Rome
Selected Publications
- 🏆 Unifying Cross-Lingual Semantic Role Labeling with Heterogeneous Linguistic Resources — NAACL 2021 Outstanding Long Paper
- Monte Carlo Temperature: a Robust Sampling Strategy for LLM’s Uncertainty Quantification Methods — TrustNLP 2025
- Handling Ontology Gaps in Semantic Parsing — *SEM 2024
- RRAML: Reinforced Retrieval Augmented Machine Learning — AIxIA 2023
- DanteLLM: Let’s Push Italian LLM Research Forward! — LREC-COLING 2024
- Generating Query Recommendations via LLMs — IR-RAG @ SIGIR 2024
Outside of work, I’m a long-time Calisthenics practitioner 💪