I founded Predates with Prof. Chris Ré, Travis Addair and Dev Rishi. Predibase makes training and serving finetuned LLMs easy, efficient and cost effective.

Working in Prof. Chris Ré's Hazy Research group on ML systems.

Worked on both research and applications in Dialogue Systems, Natural Language Processing, Deep Learning, Reinforcement Learning, Graph Learning and Meta-Learning. Created Ludwig (6.5k+ stars on GitHub), a toolbox for code-free Deep Learning, and worked on several impactful deployed applied Machine Learning projects, including NLP for Customer Support and Graph Learning for Uber Eats Recommender Systems.

Worked on dialogue systems research and applications, using Natural Language Processing, Deep Learning and Reinforcement Learning techniques.

Uber acquired Geometric Intelligence (New York Times, Wall Street Journal, Wired, TechCrunch, MIT Tech Review, BBC, Bloomberg, Fortune, Repubblica) and created Uber AI Labs, where I'm doing research about Natural Language Processing with Deep Learning. I created Uber's internal Deep Learning and NLP experimentation framework and used it in the customer support application COTA.

I've been working at Geometric Intelligence, a NYU startup led by Gary Marcus and Zoubin Ghahramani with Kenneth Stanley, Noah Goodman and Jeff Clune. We were experimenting on advances in Artificial Intelligence using Deep Learning models for hybrid language, vision and knowledge tasks. We also developed Xprop, a new algorithm for learning from less data.

I worked on several Natural Language Processing and Machine Learning projects: question understanding and classification, semantic models for cognitive search, question answering, query correction and autocomplete, text classification and knowledge induction with Deep Learning and other techniques while admiring New York's amazing skyline.

I founded QuestionCube, a startup working on question answering, and developed several question answering products: Wikiedi, a question answering system over Wikipedia with NLP and semantic technologies (in closed Beta); PArisponde, a semantic search engine over Gov FAQ Open Data (available on Apple AppStore and Google Play - Won the special prize of Apps4Italy); AQPrisponde, a question answering system for help desk operators

I worked on generative design and data visualization on the web and mobile projects at Italian design firm FF3300

I was a tutor and I've been teaching about search engines at XYLab, a workshop about new publishing, video making, hacking and network algorithms

I carried on a research project about the adoption of semantic models and network-based approaches for community question answering in the fantastic Yahoo! Lab in Barcelona

I've been tutoring and teaching the Programing Languages course of the Computer Science department of the Università degli Studi di Bari Aldo Moro

I was a journalist, head editor of the Nintendo section, international reporter, director and video editor for Everyeye.it (Hidedesign). I wrote more than 40 articles and 20 video-articles about video games

PhD in Computer Science with a thesis on semantics and question answering

I attended the European Summer School on Information Retrieval 2013 in Granada, Spain

I attended the Lisbon Machine Learning Summer School 2013 in Lisbon, Portugal

I attended the second edition of BigDive in Turin, Italy, a training program to boost the technical skills in big data analysis and visualization

I attended the Machine Learning Summer School 2012 in Kyoto, Japan

Master in Computer Science with a thesis about a question answering framework

I attended InnovAction Lab Puglia, an empirical school that teaches how to evaluate the innovative potential of an idea and how to present it to private investors and companies

Bachelor in Computer Science with a thesis about serendipity in recommender systems