Zhiyuan Lin
Research Scientist at Meta
Palo Alto, California, United States
Summary
👤
Senior🎓
Top SchoolZhiyuan Lin is a research scientist at Meta with 13 years of experience specializing in applied machine learning, Bayesian optimization, and adaptive experimentation. He holds a CS PhD from Stanford and a BS from Georgia Tech, and contributes to prominent open-source projects such as PyTorch’s BoTorch and Meta’s Ax, where he’s implemented core Pairwise Preference Learning and improved visualization for experimentation. At Meta’s Central Applied Science team he bridges research and production, turning probabilistic models and preference-learning methods into deployable experimentation tools. His background spans industry research internships at Facebook, Tencent, eBay and Yahoo, reflecting a long-running focus on recommender systems, user behavior modeling, and scalable ML. An often-overlooked strength is his hands-on work on low-level model integration (e.g., PairwiseGP and input transforms) that directly improves the fidelity of active learning and preference-based optimization.
13 years of coding experience
3 years of employment as a software developer
Bachelor’s Degree Computer Science, Bachelor’s Degree Computer Science at Georgia Institute of TechnologyDoctor of Philosophy (Ph.D.) Computer Science, Doctor of Philosophy (Ph.D.) Computer Science at Stanford University
Chinese, English
Github Skills (20)
pytorch10
python10
model-driven10
machine-learning10
gaussian-processes10
model-building10
bayesian10
optimisation10
modeling10
optimization10
model-driven-development10
data-visualizations9
apidoc9
data-visualisation9
api9
Programming languages (3)
Github contributions (5)
meta-pytorch/botorch
Nov 2019 - Oct 2022
Bayesian optimization in PyTorch
Role in this project:
Back-end Developer / ML Engineer
Back-end Developer / ML EngineerContributions:21 commits, 32 PRs, 6 comments in 2 years 11 months
Contributions summary:Zhiyuan primarily contributes to the `botorch` repository, a Bayesian optimization library in PyTorch. The commits show the user modifying the `PairwiseGP` model, implementing and adjusting features like model fitting and the use of a `ScaleKernel`. They also worked on supporting input transforms within the PairwiseGP model. Their work directly impacts core model functionality and its integration with active learning and preference learning tasks.
bayesianpytorch
facebook/Ax
Sep 2021 - Jun 2022
Adaptive Experimentation Platform
Role in this project:Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:5 commits, 65 PRs, 3 comments in 9 months
Contributions summary:Zhiyuan contributed to the Ax experimentation platform by enhancing tooltip functionality within contour plots to display parameter and metric values. They also implemented color-coding for tradeoff plots, improving the visualization of in-sample data. Furthermore, the user added support for Pairwise Preference Learning within Ax, creating `PairwiseModelBridge` and integrating it with the `RankingDataset`. Additionally, the user made modifications to support PE/PLBO by revamping the previous implementation to pass the pref_model into the EUBO acqf or serve as the objective, extracting it from a `PairwiseModelBridge`.
experimentationadaptivesimulationadaptive-experimentation-platformexperimentation-platform
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.