Jay Chae
Software Engineer at Meta
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolJay Chae is a seasoned software engineer with 12 years of experience building high-performance systems, currently focused on PyTorch performance infrastructure at Meta where he optimizes ML training scalability and profiling. He has a strong iOS engineering foundation from roles at Apple and early-stage startups, having shipped mobile apps and tooling under tight deadlines. Jay contributes to prominent open-source projects like PyTorch and Kineto, solving subtle profiling bugs and improving GPU/CPU tracing and metadata for real-world ML workloads. Known for a pragmatic "move slow, break everything" approach, he blends careful systems-level thinking with a willingness to refactor and harden complex edge cases. Based in San Francisco, he combines production-grade engineering with a track record of improving observability in large-scale ML environments.
12 years of coding experience
4 years of employment as a software developer
University of Illinois Urbana-Champaign
English, Korean
Github Skills (20)
debugging10
pytorch10
debug10
performance-monitor10
c-language10
performance-analytics10
code-profiling10
machine-learning10
performance-analysis10
gpu10
profiling10
cuda10
performance-monitoring10
cprogramming-language10
unit-testing9
Programming languages (5)
Github contributions (5)
pytorch/kineto
Feb 2021 - Sep 2022
A CPU+GPU Profiling library that provides access to timeline traces and hardware performance counters.
Role in this project:
Backend Developer
Backend DeveloperContributions:8 reviews, 42 commits, 21 PRs in 1 year 7 months
Contributions summary:Jay primarily contributed to the `kineto` project, a CPU+GPU profiling library. Their work focused on addressing and fixing issues related to the ActivityProfiler, specifically resolving a bug where the `ActivityProfiler` was writing to a singleton with a different reference than the `ActivityProfilerController`. They also refactored code, handled edge cases in timestamps, and contributed to the addition of JSON-formatted metadata for client-side activities. The user also updated manual Kineto runs.
cudacputracesprofilingtimeline
pytorch/pytorch
Apr 2021 - Oct 2022
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:ML Engineer
ML EngineerContributions:66 reviews, 24 commits, 21 PRs in 1 year 6 months
Contributions summary:Jay's commits primarily focus on the Kineto profiling tool within the PyTorch framework. They deprecated and removed outdated components related to Kineto, ensuring it's the primary profiling option. Further contributions include optimizing Kineto integration with the PyTorch profiler and addressing issues in on-demand tracing, specifically related to input shape collection. The commits demonstrate a focus on improving profiling capabilities within the machine learning framework.
pythongpu-accelerationdeep-learninggpunumpy
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.