Sara Piscitelli

Rome, Italy ·

AI Research Engineer @ Leonardo Labs

AI Research Engineer with a Master’s degree in Computer Engineering, specializing in Natural Language Processing (NLP) applications.
Extensive experience in developing software and AI models, with a primary focus on the NLP field.
Professional interests encompass Natural Language Processing, Reinforcement Learning, and Software Engineering.

Socials

Experience

AI Research Engineer

Leonardo Company (Leonardo Labs)

Engaged in research and software engineering activities for both internal and European projects.
Engineered software systems using NLP models for automated information extraction from documents.
Applied large language models (LLMs) across multiple applications (e.g. question answering, clustering, text generation, triplet extraction, …).
Worked on image super-resolution employing Generative Adversarial Networks (GANs).

May 2021 - Present

Junior Data Scientist

LINKS Foundation

Involved in research and software engineering activities for European projects.
Contributed to the design, implementation, and ongoing maintenance of a distributed system analyzing social media.
Applied NLP methods for analyzing and processing textual data sources.

April 2020 - May 2021

Education

University of Rome, Tor Vergata

Master’s degree in Computer Engineering

Data Science and Engineering specialization

Thesis: A new agent-based NAS (Neural Architecture Search).
Final mark: 110/110 cum laude

Oct. 2017 - Feb. 2020

University of Rome, Tor Vergata

Bachelor’s degree in Computer Engineering

Final mark: 110/110 cum laude

Sept. 2014 - Oct. 2017

Skills

Artificial Intelligence

Natural Language Processing
Reinforcement Learning
Deep Learning
Machine Learning

Libraries & Frameworks

PyTorch
HuggingFace Libraries (transformers, datasets, peft)
LangChain
Semantic Kernel
Pandas
Numpy
Asyncio & Multiprocessing

Languages, Operating Systems & Tools

Python
git
linux
bash

Data Management

MySQL
MongoDB
Neo4j
Redis

Deployment

Docker
Microservices Architecture
RESTful APIs Architecture

Publications

Workshop - An adaptable method for developing an Open-domain Question Answering system

Ital-IA 2023: 3rd National Conference on Artificial Intelligence. In the AI for industry workshop, we presented an open-domain question-answering system that retrieves documents from a knowledge corpus and extracts answers using a fine-tuned language model, employing a 'zero-shot' approach.

May 2023

Conference - Multilingual Text Classification from Twitter during Emergencies

2021 IEEE International Conference on Consumer Electronics (ICCE). At the conference, we demonstrated a multilingual tool that categorizes tweet content using a deep learning classifier with multilingual word embeddings. The model, initially trained in one language, efficiently adapts to multiple languages via zero-shot inference.

January 2021

Projects

Natural Language Processing Tasks

The repository contains a series of natural language processing (NLP) tasks addressed using various methods and models. The tasks include text classification, named entity recognition, question answering, summarization, and text generation.

Web Data Extraction

A simple web crawler and scraper has been developed to extract data from websites. The crawler operates on Scrapy, while the scraper utilizes BeautifulSoup. At present, the scraper is configured to extract structured data specifically from the Trustpilot.com website. For other sites, the scraper is designed to return unstructured text data