Ehsan Amid
Research Scientist at Google DeepMind
San Francisco, California, United States
Summary
👤
Senior🎓
Top SchoolEhsan Amid is a machine learning researcher-engineer with 11 years of experience, currently a Member of Technical Staff at Core Automation and a Research Scientist at Google DeepMind contributing to Gemini. He holds a PhD in Computer Science from UC Santa Cruz and an MSc in Machine Learning and Data Mining from Aalto University, with early foundations in telecommunications engineering. His work spans theory and practice—tempered Bregman divergences, robust loss functions (e.g., bi-tempered logistic loss), optimization and dimensionality reduction techniques like TriMap—bridging principled math with scalable implementations in TensorFlow/JAX. At Google Brain he developed robust training objectives, adaptive optimizers and block-sparse training methods, and his open-source contributions to the prominent google-research repo reflect deep expertise in loss design and numerical computation. Based in San Francisco, he combines academic rigor with production-savvy engineering across research and applied ML. A less obvious strength is his consistent focus on robustness and optimization that links online learning theory to practical, highly testable model training tools.
11 years of coding experience
5 years of employment as a software developer
Master's degree Machine Learning and Data Mining, Master's degree Machine Learning and Data Mining at Aalto UniversityAmirkabir University of TechnologyUniversity of California Santa Cruz
Azerbaijani, Turkish, Persian, English, Spanish, Arabic
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (10)
numerics10
machine-learning10
loss-functions10
tensorflow10
jax10
numerical10
computation10
numeric10
ai10
python10
Programming languages (2)
Github contributions (5)
google-research/google-research
Jul 2019 - Aug 2022
Google Research
Role in this project:
ML Engineer
ML EngineerContributions:14 commits in 3 years 1 month
Contributions summary:Ehsan primarily contributed to the implementation and testing of a bi-tempered logistic loss function within a TensorFlow/JAX environment, indicating a focus on machine learning model training. Their work included defining functions for tempered softmax and sigmoid, and supporting both sparse and general bi-tempered logistic loss calculations, highlighting expertise in loss functions. Further contributions included incorporating JAX for compatibility. This indicates expertise in machine learning, numerical computation, and a specific interest in loss function design and optimization.
googlemachine-learningai
eamid/trimap
Apr 2018 - Aug 2022
TriMap: Large-scale Dimensionality Reduction Using Triplets
Contributions:75 commits, 3 PRs, 86 pushes in 4 years 4 months
pythondimensionalitydimensionality-reductiontripletsfeature-extraction
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.