Chien-chin Huang
Research Scientist at Facebook
Cupertino, California, United States
Summary
🤩
Rockstar🎓
Top SchoolChien-chin Huang is a research scientist in Cupertino with 12 years of experience building and optimizing large-scale systems for ML and recommendation workloads. At Facebook he focuses on distributed training and efficiency, contributing upstream to PyTorch—improving FSDP, allreduce primitives, and distributed checkpointing to better support mixed-precision and memory-constrained models. His background blends academic rigor (PhD in Computer Science from NYU) with industry practice from roles at MediaTek, Google, IBM and academic research groups, giving him deep systems and performance expertise. He has hands-on experience hardening metrics and throughput logic in torchrec, demonstrating attention to correctness and production-readiness in domain libraries. Notably, his work often targets subtle failure modes (optimizer state handling, division-by-zero risks) that improve reliability at scale.
12 years of coding experience
11 years of employment as a software developer
Doctor of Philosophy (Ph.D.) Computer Science, Doctor of Philosophy (Ph.D.) Computer Science at New York UniversityMaster's Degree Computer Science, Master's Degree Computer Science at National Tsing Hua University
Chinese, English
Github Skills (12)
pytorch10
distributed-training10
recommendation-system10
python10
metric10
tensorflow9
performance-optimization9
operation9
deeplearning-ai9
deep-learning9
tensorrt9
tensor9
Programming languages (3)
Github contributions (5)
pytorch/pytorch
Feb 2022 - Jan 2023
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
Back-end Developer
Back-end DeveloperContributions:922 reviews, 482 commits, 368 PRs in 11 months
Contributions summary:Chien-chin's contributions focused on enhancing the PyTorch framework, specifically in the realm of distributed training with FSDP (Fully Sharded Data Parallelism). Their work involved addressing and resolving issues related to optimizer state dicts within the DDP system, ensuring compatibility with models using mixed precision and handling potential memory inefficiencies. They also implemented and optimized allreduce and other communication primitives, improving the performance and usability of distributed checkpointing for various workloads.
pythongpu-accelerationdeep-learninggpunumpy
pytorch/torchrec
Apr 2022 - Aug 2022
Pytorch domain library for recommendation systems
Role in this project:Back-end Developer
Back-end DeveloperContributions:1 review, 12 commits, 11 PRs in 4 months
Contributions summary:Chien-chin contributed to the torchrec library by addressing various issues related to metrics and throughput calculations. They refactored existing code, removing deprecated features and addressing potential errors like division by zero. Furthermore, the user updated the metrics module by incorporating new features such as adaptive compute intervals and corrected typing issues, improving the library's overall functionality.
cudapytorchrecommendation-systemsdeep-learninggpu
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.