Johnny Liu
California, United States
Summary
🤩
Rockstar🎓
Top SchoolJohnny Liu is an engineering leader and former SDE/MLE with 11+ years of hands-on experience building large-scale ML, HPC, and data infrastructure, most recently leading the Rufus LLM platform for Amazon Shopping. He combines PhD-level expertise in parallel I/O and big data with production ML systems work—driving foundation-model training, agentic RL, and massive data pipelines for internet and customer behavior signals. At TikTok he architected cloud-native ML platforms and auto-tuning systems that cut training and processing times dramatically, and his open-source contributions include MPI-aware parallel I/O work in the widely used h5py HDF5 Python library. A long-standing NSF and DOE panelist since 2015, he evaluates multi-million-dollar proposals bridging research breakthroughs to real-world impact. He has repeatedly translated HPC research into production—co-founding an AIoT startup and shipping edge/cloud hybrid ML systems used by 100k+ customers—showing a rare blend of research depth and product execution. Based in California, he prefers tackling system-level bottlenecks where algorithms, storage, and compute meet.
11 years of coding experience
13 years of employment as a software developer
Doctor of Philosophy - PhD, Computer Science, Doctor of Philosophy - PhD, Computer Science at Texas Tech University
Github Skills (9)
data-storage10
hdf10
mpi10
parallel-computing10
python10
numpy9
cython8
back-end-development8
devops7
Programming languages (8)
Github contributions (5)
h5py/h5py
Nov 2015 - Dec 2015
HDF5 for Python -- The h5py package is a Pythonic interface to the HDF5 binary data format.
Role in this project:
Back-end Developer & DevOps Engineer
Back-end Developer & DevOps EngineerContributions:12 commits, 4 PRs, 9 comments in 28 days
Contributions summary:Johnny's contributions primarily involved implementing and testing features related to parallel I/O support in the h5py library, specifically focusing on MPI integration for collective reads and writes. They added context managers and API calls to manage MPI transfer modes, and they included test code to validate the collective I/O functionality. Furthermore, the user modified existing code to align with MPI environments, while also refactoring the code.
h5pybinary-datapythondata-formathdf5
NERSC/object-store
Jan 2018 - Dec 2018
Contributions:56 commits, 38 pushes, 1 issue in 11 months
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.