Wushi Dong
Menlo Park, California, United States
Summary
👤
SeniorWushi Dong is a software engineer with eight years of experience specializing in LLM inference systems, runtime performance, and distributed ML at scale. Currently at Meta, he focuses on compiler runtime and efficiency for the Meta Training and Inference Accelerator, following work at AWS where he led ML compiler integrations and efficient partitioning of large LLMs for SageMaker. His background includes a Physics Ph.D. from the University of Chicago and HPC-scale research—he helped scale Flood-Filling Network training to 2048 KNL nodes (131,072 cores) for connectomics. Wushi blends deep academic rigor with production-focused compiler and systems engineering, and has a history of open-source tooling from his MPI C++ simulation library to ML repos released during an IBM internship. Based in Menlo Park, he brings a rare combination of low-level performance optimization and practical deployment experience for AI hardware and cloud.
8 years of coding experience
Github Skills (12)
cuda10
llama10
keras10
pytorch10
transformer10
gpt10
inference10
openai10
action-recognition10
tpu10
amd10
llm10
Programming languages (2)
Github contributions (5)
wushidonguc/distributed_ffn
Jan 2019 - Feb 2020
Contributions:21 commits, 19 pushes, 1 branch in 1 year 1 month
wushidonguc/two-stream-action-recognition-keras
Mar 2018 - Aug 2019
Two-stream CNNs for video action recognition implemented in Keras
Contributions:1 release, 26 commits, 23 pushes in 1 year 4 months
action-recognitionkerasvideotwo-stream-cnnucf-101
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.