Giovanni Puccetti
Researcher at IIT-CNR
Tuscany, Italy
Summary
👤
Senior🎓
Top SchoolGiovanni Puccetti is a researcher in NLP and language modeling with eight years of experience spanning academic and applied roles at CNR (IIT and ISTI) and a PhD in Data Science from Scuola Normale Superiore. He transitioned from a mathematical foundation—MSc in Mathematics from the University of Amsterdam—into hands-on ML engineering, notably contributing to open-source CLIP implementations and refining the CoCa model’s generation, beam search, and training losses. Based in Tuscany, he blends rigorous theoretical training with practical system-level fixes, such as addressing accumulative gradient issues and integrating Hugging Face tooling. Giovanni’s profile reflects a researcher comfortable shipping production-ready model components while remaining engaged in cutting-edge language-model research.
8 years of coding experience
6 years of employment as a software developer
Master's degree, Mathematics, Master's degree, Mathematics at University of AmsterdamLaurea triennale, Matematica, Laurea triennale, Matematica at Università di Pisa
Stackoverflow
293reputation
8kreached
13answers
1question
Github Skills (14)
transformers10
computer-vision10
pytorch10
machine-learning10
language-model10
python10
deep-learning9
pre-trained-model7
macros6
struct6
multiple-dispatch6
julia6
linear-algebra6
jupyter6
Programming languages (4)
Github contributions (5)
mlfoundations/open_clip
Dec 2022 - Jan 2023
An open source implementation of CLIP.
Role in this project:
ML Engineer
ML EngineerContributions:14 reviews, 16 commits, 40 PRs in 1 month
Contributions summary:Giovanni significantly contributed to the CoCa (Contrastive Captions) model within the repository. They implemented generation and beam search functionalities, including necessary utilities and integrations with Hugging Face transformers. The user refined the CoCa model's training, addressing accumulative gradient issues, and modified the loss function. Moreover, the user improved the generation process by integrating various features.
pytorchlanguage-modelzero-shot-classificationdeep-learningclip
gpucce/open_clip
Nov 2022 - Jan 2025
An open source implementation of CLIP.
Contributions:3 PRs, 325 pushes, 46 branches in 2 years 2 months
deep-learningclip
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.