Anurag Arnab
Research Scientist at Google
Grenoble, Auvergne-Rhône-Alpes, France
Summary
🤩
Rockstar🎓
Top SchoolAnurag Arnab is a research scientist at Google with 12 years of experience bridging deep learning research and production-quality systems, focused on computer vision and model training. He completed a PhD at the University of Oxford’s Torr Vision Group and has published and interned at leading labs including DeepMind, with work spanning semantic segmentation, structured prediction and large-scale image retrieval (contributions that fed into Google Lens). At Google and in prominent open-source projects like Scenic and Flax he has improved training pipelines, augmentation strategies and checkpointing robustness for JAX/Flax-based vision models. Comfortable moving ideas from research prototypes to reliable code, he brings both academic rigor and practical engineering, including hands-on fixes that improve preemption-safe checkpointing and ViT/ImageNet configurations. Based in Grenoble, France, he pairs strong mentoring and teaching roots with a track record of shipping scalable ML components in production.
12 years of coding experience
4 years of employment as a software developer
B.Sc (Engineering), Electrical and Computer, B.Sc (Engineering), Electrical and Computer at University of Cape TownDoctor of Philosophy (Ph.D.), Computer Vision, Doctor of Philosophy (Ph.D.), Computer Vision at University of OxfordCourseraWestville Boys' High School
English, Afrikaans, Bengali
Github Skills (10)
transformers10
computer-vision10
error-handling10
deeplearning-ai10
deep-learning10
jax10
python10
flax10
testing9
machine-learning9
Programming languages (6)
Github contributions (5)
google-research/scenic
Jul 2021 - Nov 2022
Scenic: A Jax Library for Computer Vision Research and Beyond
Role in this project:
ML Engineer
ML EngineerContributions:59 commits, 3 comments in 1 year 4 months
Contributions summary:Anurag primarily contributed to the training and configuration aspects of the "Scenic" library, which focuses on computer vision research. Their work included adding a mixup augmentation technique to a classification trainer and correcting configuration files for Vision Transformers (ViT) on the ImageNet dataset. They also made internal changes, indicating involvement in the project's internal workings. The commits demonstrate a focus on model training and optimization within a computer vision context.
computer-visionjaxdeep-learningtransformersvision-transformer
google/flax
Jan 2021 - Apr 2022
Flax is a neural network library for JAX that is designed for flexibility.
Role in this project:Back-end Developer
Back-end DeveloperContributions:1 review, 6 commits, 4 PRs in 1 year 3 months
Contributions summary:Anurag focused on improving the checkpointing functionality within the Flax library. Their contributions included fixing a bug related to preemption during checkpoint saving, adding comments to clarify the purpose of specific checks, and correcting documentation errors. They also made a minor fix to the device handling in the jax_utils module. These changes primarily centered on enhancing the reliability and robustness of the library's core features.
jaxneural-network
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.