Ofir Zafrir
Deep Learning Applied Research at Intel Corporation
Israel
Summary
🤩
Rockstar🎓
Top SchoolOfir Zafrir is an applied deep learning researcher at Intel Labs with a decade of engineering experience and a BS in Computer Engineering from Technion. He specializes in NLP model optimization and quantization, having contributed quantization-aware training, quantized embeddings and linear layers, and test coverage for quantized BERTs in the widely used IntelLabs/nlp-architect library. At Intel he progressed from intern to research engineer, bridging state-of-the-art models with hardware-aware algorithms to make production-ready NLP more efficient. Earlier embedded-systems work at Rafael gave him practical real-time C and MATLAB skills that inform his system-minded approach to model deployment.
10 years of coding experience
2 years of employment as a software developer
Bachelor of Science - BS, Computer Engineering, Bachelor of Science - BS, Computer Engineering at Technion - Israel Institute of Technology
Hebrew, English
Github Skills (10)
transformers10
quantization10
pytorch10
nlp10
deep-learning10
bert10
classify9
classification9
text-classification9
sequence-labeling9
Programming languages (3)
Github contributions (5)
IntelLabs/nlp-architect
Nov 2018 - Dec 2019
A model library for exploring state-of-the-art deep learning topologies and techniques for optimizing Natural Language Processing neural networks
Role in this project:
ML Engineer
ML EngineerContributions:25 commits, 10 PRs, 4 pushes in 1 year 2 months
Contributions summary:Ofir primarily focused on integrating and developing quantized BERT models within the repository, specifically for tasks like sequence classification and token classification. Their contributions include implementing quantization-aware training, building modules for quantized embeddings and linear layers, and creating specific quantized BERT models. These changes included modifying the BERT model architecture and adding test cases to validate the quantization process. The user also worked on documentation related to the quantized BERT implementation.
nlunatural-language-understandingbertlanguage-processingstate-of-the-art
IntelLabs/Model-Compression-Research-Package
Nov 2021 - Aug 2022
A library for researching neural networks compression and acceleration methods.
Contributions:10 commits, 19 PRs, 15 pushes in 8 months
deep-learningpruningcompressionaccelerationneural-networks
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.