Valerio Sofi
Machine Learning Engineer at Prima
Turin, Piedmont, Italy
Summary
🤩
Rockstar🎓
Top SchoolValerio Sofi is a Machine Learning Engineer from Turin with nine years of experience building and optimizing AI systems across computer vision and NLP. He has trained and fine-tuned large language models to extract actionable insights for product analytics at Nebuly and deeply focused on model optimization using advanced compilers and quantization techniques. Previously he developed CNN-based identity-document and face analysis microservices (including OCR) for remote onboarding at Intesa/IBM, and earlier built hybrid mobile apps during his internship. Valerio contributes to open-source tooling for model performance (nebuly-ai/optimate), adding PyTorch-TensorRT support, quantization, and cross-library fixes that improved deployment reliability. He holds a Laurea Magistrale in Computer Engineering from Politecnico di Torino and combines research-grade AI skills with hands-on production deployment experience. Notably, his work bridges model training and low-level performance engineering, making models both smarter and faster in real-world systems.
9 years of coding experience
6 years of employment as a software developer
Maturità scientifica, 97/100, Maturità scientifica, 97/100 at Liceo Scientifico Statale Galileo FerrarisLaurea Magistrale, Computer Engineering, 110, Laurea Magistrale, Computer Engineering, 110 at Politecnico di Torino
English, Italian
Github Skills (10)
tensorrt10
pytorch10
machine-learning10
deep-q-learning10
deep-learning10
ai10
model-optimization10
optimization10
onnx9
large-language-models8
Programming languages (6)
Github contributions (5)
nebuly-ai/optimate
Jun 2022 - Jan 2023
A collection of libraries to optimise AI model performances
Role in this project:
ML Engineer
ML EngineerContributions:63 reviews, 42 commits, 47 PRs in 7 months
Contributions summary:Valerio primarily worked on optimizing AI model performance within the `nebuly-ai/optimate` repository. Their contributions included addressing requirements conflicts, auto-deleting temporary files, and resolving various installation and versioning issues related to libraries like NumPy, SciPy, and OpenVINO, as well as fixing typos and implementing several tests. They also introduced PyTorch-TensorRT support by addressing autoinstaller issues, adding quantization, and fixing bugs related to half precision.
pypicompilertensorflowtransformersedge-computing
valeriosofi/flair
Sep 2022 - Sep 2022
A very simple framework for state-of-the-art Natural Language Processing (NLP)
Contributions:27 pushes, 1 branch in 11 days
nlpartbertsimple-frameworklanguage-processing
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.