Eirikur Agustsson
Senior Research Scientist at Google DeepMind
Zurich, Zurich, Iceland
Summary
👤
Senior🎓
Top SchoolEirikur Agustsson is a Senior Research Scientist with nine years of experience at Google and Google DeepMind, specializing in large-scale pretraining and machine learning for compression. He holds a PhD from ETH Zürich's Computer Vision Lab and a strong technical foundation in electrical engineering and mathematics. His research and engineering work spans model architecture, training, and optimization, with notable open-source contributions to tensorflow/compression where he improved entropy models, soft rounding, and the HiFiC codec. Based in Zurich and originally from Iceland, he bridges rigorous academic research with production-grade ML engineering. Colleagues know him for refactoring complex model components into reusable libraries and for making state-of-the-art compression methods more practical. He combines deep theoretical knowledge with hands-on coding to push both research boundaries and real-world system performance.
9 years of coding experience
9 years of employment as a software developer
BS Mathematics, BS Mathematics at Háskóli ÍslandsICVSS 2015Doctor of Philosophy - PhD Computer Vision Lab, Doctor of Philosophy - PhD Computer Vision Lab at ETH Zürich
Icelandic, English, German
Github Skills (12)
neural-network10
model-building10
compression10
machine-learning10
deep-learning10
tensorflow10
python10
modeling10
compress10
model-driven10
model-driven-development10
distributions9
Programming languages (5)
Github contributions (5)
tensorflow/compression
Jun 2020 - Jul 2022
Data compression in TensorFlow
Role in this project:
ML Engineer
ML EngineerContributions:2 reviews, 8 commits, 1 comment in 2 years 1 month
Contributions summary:Eirikur primarily contributed to the development and improvement of machine learning models within the TensorFlow Compression library. Their work included splitting and refactoring distribution classes, such as DeepFactorized, and adding new functionality like the universally quantized entropy model. The user also integrated soft rounding operations and layers and made enhancements to the HiFiC model, improving its usability and performance. These contributions suggest a focus on model architecture, training, and optimization within the context of data compression.
pythondeep-learningcompressionmachine-learningneural-network
relational/relational.github.io
Jul 2017 - Jul 2017
Contributions:9 pushes, 1 branch in 3 days
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.