Abin Shahab
Senior Staff Engineer at Apple
Mountain View, California, United States
Summary
🤩
Rockstar🎓
Top SchoolAbin Shahab is a Senior Staff Engineer with 9 years focused on deep learning infrastructure and a long track record building scalable distributed systems for companies like Apple and LinkedIn. He led LinkedIn’s transition from single-node training to managed, multi-tenant large-scale training clusters (hundreds of nodes, thousands of GPUs), productionized federated and privacy-preserving ML, and helped bring Kubernetes into LinkedIn’s stack. An active open-source maintainer and contributor to Horovod, he’s fixed race conditions, enhanced Ray integration, and added elastic job support to a widely used distributed training framework. Based in Mountain View, he now builds ML infra powering AI on millions of Apple devices, blending low-level performance work (TensorFlow I/O, C++ optimizations) with platform and orchestration engineering.
9 years of coding experience
18 years of employment as a software developer
The University of ArizonaMasters Software Engineering, Masters Software Engineering at Carnegie Mellon University
Github Skills (11)
ray10
tensorflow10
devops10
python10
horovod10
dockers9
cicd9
docker9
keras9
machine-learning8
mpi6
Programming languages (7)
Github contributions (5)
horovod/horovod
Oct 2021 - Jan 2022
Distributed training framework for TensorFlow, Keras, PyTorch, and Apache MXNet.
Role in this project:
DevOps Engineer & ML Engineer
DevOps Engineer & ML EngineerContributions:27 reviews, 6 commits, 4 PRs in 3 months
Contributions summary:Abin contributed to the improvement of the Ray integration for the Horovod framework by fixing race conditions in the elastic scaling tests. They modified the Dockerfile to use a specific Ray version for testing and addressed gradient aggregation issues within TensorFlow 2 Keras tests, ensuring the use of real gradients. The user further enhanced the framework by adding support for elastic and static jobs in the RayExecutor API. Finally, the user refactored the code, replacing `tempfile.mktemp()` with `tempfile.mkstemp()` and added support for resurrecting blacklisted hosts.
mpikeras-tensorflowtrainingbaidutensorflow
ashahab/tensorflow
Sep 2019 - May 2021
An Open Source Machine Learning Framework for Everyone
Contributions:44 pushes, 4 branches in 1 year 8 months
pythondata-sciencedeep-learningmachine-learningframework-learning
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.