Shaden Smith
Member Of Technical Staff at Microsoft AI
Bellevue, Washington, United States
Summary
🤩
Rockstar🎓
Top SchoolShaden Smith is a Member of Technical Staff at Microsoft AI with 14 years of experience building high-performance algorithms and scalable distributed systems for machine learning and data mining. She drives large-scale deep learning infrastructure work on DeepSpeed, contributing integration, checkpointing, and debugging enhancements used to train transformer models like BERT and GPT-class workloads. Her academic roots produced SPLATT, an open-source tensor factorization toolkit that has scaled to over 16,000 cores and is adopted across academia, industry, and government. Shaden blends research rigor from a PhD in computer science with hands-on engineering at organizations including Intel Labs and Microsoft, delivering both corescale performance and practical ML tooling. She has a track record of shipping reproducible examples and tests (DeepSpeedExamples, Megatron-DeepSpeed) that make cutting-edge distributed training accessible. Based in Bellevue, WA, she’s passionate about squeezing performance out of code while keeping developer experience and debuggability front and center.
14 years of coding experience
12 years of employment as a software developer
Doctor of Philosophy (PhD) Computer Science, Doctor of Philosophy (PhD) Computer Science at University of MinnesotaBachelor of Science (B.S.) Computer Science, Bachelor of Science (B.S.) Computer Science at University of Kentucky
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (23)
pytorch10
distributed-training10
checkpoint10
pytest10
python10
machine-learning-models10
testing10
machine-learning10
checkpointing10
transformer-models10
deepspeed10
gpu10
bert10
parallelization9
git-repository9
Programming languages (12)
Github contributions (5)
deepspeedai/DeepSpeed
Jan 2020 - Nov 2021
DeepSpeed is a deep learning optimization library that makes distributed training and inference easy, efficient, and effective.
Role in this project:
ML Engineer
ML EngineerContributions:111 reviews, 135 commits, 211 PRs in 1 year 9 months
Contributions summary:Shaden's contributions primarily revolve around modifying and testing DeepSpeed's functionalities related to model execution and distributed training. The user added scripts and configurations to run DeepSpeed with the BingBertSquad model. They implemented distributed testing using pytest and introduced enhancements for activation checkpointing. Furthermore, the user addressed several bug fixes within the framework.
billion-parametersfinetuningtrainingmixture-of-expertszero
deepspeedai/DeepSpeedExamples
Feb 2020 - Mar 2021
Example models using DeepSpeed
Role in this project:ML Engineer
ML EngineerContributions:4 reviews, 15 commits, 13 PRs in 1 year 1 month
Contributions summary:Shaden primarily contributes to examples leveraging DeepSpeed for model training. Their work includes fixing Apex calls and integrating BERT models, specifically addressing issues within the `nvidia_run_squad_baseline.py` file. Further contributions involve a BERT example implementation, demonstrating familiarity with DeepSpeed training pipelines for transformer models. Also, this user added a pipeline parallelism example for AlexNet.
deep-learningpytorchdeepspeed
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.