Pengchong Jin
Research Scientist at Apple
Mountain View, California, United States
Summary
🤩
Rockstar🎓
Top SchoolPengchong Jin is a research scientist and engineering leader with 12 years of experience building production-grade computer vision and generative AI systems, currently working on multimodal foundation models at Apple in Mountain View. He has led large teams at Google to deliver generative image and video products for enterprise customers and built core detection frameworks and state-of-the-art CV models (SpineNet, BigNAS, COMISR) used across Google services. His hands-on background spans ML systems, model efficiency, and encoding optimizations—evidenced by contributions to high-profile open-source projects like TensorFlow TPU/models and VP9 encoder work in libvpx. Comfortable bridging research and product, he repeatedly turns novel model ideas into scalable infrastructure and customer-facing APIs. An unexpected thread through his career is deep practical work on first-pass encoding and mask-RCNN internals, showing a knack for squeezing performance at both algorithmic and systems levels.
11 years of coding experience
11 years of employment as a software developer
Doctor of Philosophy (Ph.D.), Doctor of Philosophy (Ph.D.) at Purdue UniversityHong Kong University of Science and Technology (HKUST)
Github Skills (32)
algorithm10
code-optimization10
vp910
pytorch10
algorithms10
video-encoding10
python10
encoder10
machine-learning10
c1110
ml10
c1710
data-preprocessing10
mask-rcnn10
mle10
Programming languages (4)
Github contributions (5)
google/gemma_pytorch
Feb 2024 - Jan 2025
The official PyTorch implementation of Google's Gemma models
Role in this project:
ML Engineer
ML EngineerContributions:16 PRs, 23 pushes, 55 comments in 10 months
Contributions summary:Pengchong contributed to the development of Google's Gemma models in PyTorch. Their work involved fixing a bug related to a quantification flag, modifying the activation function to use `approximate=tanh` for GeLU, and exposing logits in the `Sampler` and `GemmaForCausalLM` modules. These changes demonstrate an understanding of the model architecture and how it interacts with the sampling process.
gemmagooglepytorch
tensorflow/tpu
Dec 2018 - Oct 2020
Reference models and tools for Cloud TPUs.
Role in this project:ML Engineer
ML EngineerContributions:186 commits, 29 PRs, 44 pushes in 1 year 10 months
Contributions summary:Pengchong's commits primarily involve modifications to various components within a machine-learning related project, particularly focusing on the Mask R-CNN model. The changes include updates to data loading and parsing, implementing faster-rcnn image preprocessing, and modifications to the loss functions. Their contributions appear to enhance the model's functionality and performance in object detection and segmentation tasks.
cloud
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.