Matteo Pagliardini
Research Intern at Meta
Lausanne, Vaud, Switzerland
Summary
🤩
Rockstar🎓
Top SchoolMatteo Pagliardini is a machine learning researcher and engineer with 12 years of experience, currently pursuing a PhD at EPFL and conducting ML research at Meta FAIR. He specializes in language modeling, Transformer-based NLP, domain adaptation and robustness under distribution shift, with a track record of internships at Apple and Berkeley that bridge theory and applied research. Matteo contributed core enhancements to the widely used sent2vec project—adding training-method innovations and multithreaded Cython inference—demonstrating both deep algorithmic understanding and systems-level implementation skills. He has experience shipping ML systems in industry roles at Iprova and earlier research in computer vision and user-behavior modeling, reflecting a pragmatic bent toward scalable, efficient solutions. Based in Lausanne, he focuses on democratizing large language models and improving uncertainty estimation in deep models.
12 years of coding experience
5 years of employment as a software developer
Ecole polytechnique fédérale de LausanneBachelor's degree, Computer Science and Electronics, Bachelor's degree, Computer Science and Electronics at Ecole supérieure de Chimie Physique Electronique de Lyon
French, English, Spanish, Japanese
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (14)
machine-learning10
nlp10
trainings10
modeling10
cython10
fasttext10
cprogramming-language9
c-language9
continuous-deployment8
ml-deployment8
multithreading8
apidoc7
api7
python6
Programming languages (4)
Github contributions (5)
epfml/sent2vec
Mar 2017 - Aug 2022
General purpose unsupervised sentence representations
Role in this project:
ML Engineer
ML EngineerContributions:43 commits, 13 PRs, 37 pushes in 5 years 5 months
Contributions summary:Matteo appears to be involved in developing and improving the sent2vec model. Their commits include core modifications to the fasttext library, particularly focusing on the sent2vec training method. These changes involve adjustments to the core training loop, the addition of a cbow-c+w-ngrams method, and the implementation of multithreaded inference in a Cython wrapper. The user's work also includes updates to the model saving and loading functionalities, and the creation of a Cython wrapper for inference.
nlpsentencerepresentationsword-embeddingsunsupervised
epfml/llm-baselines
Mar 2023 - Mar 2025
nanoGPT-like codebase for LLM training
Contributions:1 review, 10 PRs, 27 pushes in 1 year 11 months
llmspretraining
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.