Heinrich Jiang
Research Scientist at Google
Mountain View, California, United States
Summary
👤
Senior🎓
Top SchoolHeinrich Jiang is a Research Scientist at Google with 13 years of experience applying machine learning and quantitative methods to real-world problems, from cloud networking to high-frequency trading. He focuses on large language models, data labeling and noisy-label robustness, constrained optimization, fairness, and model interpretability, and has driven R&D that spans both theory and production. A summa cum laude Princeton mathematician, he combines rigorous mathematical training with hands-on systems experience gained across Google engineering roles and quantitative trading desks. Notably, his background in creating high-frequency trading signals informs a pragmatic, data-driven approach to uncertainty estimation and active learning in ML systems. Based in Mountain View, he blends research-grade publications with product-minded development to move advances from prototype to scale.
13 years of coding experience
2 years of employment as a software developer
A.B., Mathematics, summa cum laude, A.B., Mathematics, summa cum laude at Princeton University
English, Russian, Ukrainian
Github Skills (22)
classifier10
softmax10
shift9
probability9
uncertainty9
trust9
arxiv8
points8
machine-learning7
mean-shift7
act6
clustering-algorithm6
clustering5
neural-network5
singleton5
Programming languages (4)
Github contributions (5)
google/TrustScore
Sep 2018 - Nov 2021
To Trust Or Not To Trust A Classifier. A measure of uncertainty for any trained (possibly black-box) classifier which is more effective than the classifier's own implied confidence (e.g. softmax probability for a neural network).
Contributions:2 reviews, 3 commits, 1 PR in 3 years 2 months
confidenceclassifierboxeffectivemachine-learning
google/quickshift
Mar 2018 - Aug 2018
A clustering algorithm that first finds the high-density regions (cluster-cores) of the data and then clusters the remaining points by hill-climbing. Such seedings act as more stable and expressive cluster-cores than the singleton modes found by popular algorithm such as mean shift. (https://arxiv.org/abs/1805.07909)
Contributions:2 commits, 1 PR, 2 pushes in 4 months
shiftk-meansclustersdata-miningmodes
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.