I work at Meta as an ML Research Scientist, currently in the Meta AI Org in Menlo Park, California. I joined Meta in August 2025, after finishing my Ph.D. at Purdue, where I studied bias and fairness in AI: whether AI treats people fairly, and how people react to it and use it.
Ph.D. in Computer Science, Purdue University · completed June 2025
My PhD followed the machine learning pipeline from the data people create, through the models that data shapes, to the model decisions people experience. At Meta I build that pipeline itself, for real users, from ranking what people see to the data models learn from.
My experience at Meta and in my Ph.D. can be summarized by this loop.
People create data. A model learns from it and makes decisions about people. People react to the decisions made about them, and their reactions are collected as data and used to update the model again, so the cycle starts again.
I help AI systems become better (SOTA). At Meta AI, I work on the data that AI models learn from: improving the data we already have, and creating new data for models to learn from. I also build the step-by-step tools and workflows that make this work possible, and evaluate AI agent systems.
Computer Science, completed June 2025. Advised by Dr. Ming Yin. This loop became my dissertation. Its four papers, generated from this loop, are in the next section.
Purdue UniversityPh.D. Dissertation · Department of Computer Science · June 2025The chapters of the dissertation are the four papers in this section, in the order they connect to each other. Advised by Dr. Ming Yin.
Theme · Crowdsourcing — labels from many people
Quick recap: When labels come from many people, each person's confirmation bias comes with them — the tendency to favor what they already believe. The model in this paper estimates each annotator's bias and corrects for it while combining the labels, so the result is substantially less biased.
The 30th International Joint Conference on Artificial Intelligence (IJCAI), Online, August 2021.
Themes · Fairness · People and AI
Quick recap: When people face an AI's decisions again and again — here, loan decisions — what shapes their fairness judgments, and whether they stay or leave, is less whether the system treats every group the same and more whether it favors their own group. For the people a system decides about, fairness is judged from where their own group stands.
The 5th AAAI/ACM Conference on AI, Ethics, and Society (AIES), Oxford, UK, August 2022.
Themes · Fairness · People and AI
Quick recap: Across three experiments, when people had a real chance to improve their qualifications, an AI's unfairness did not reduce their engagement or their effort to improve. But it did not soften their judgment either: when the AI treated their own group worse than the others, people still saw the system as less fair.
Themes · Fairness · People and AI
Quick recap: When an AI hands out tasks and keeps updating who gets what, people are not reacting to a single decision but to a moving series of them. This chapter follows how people respond as the allocations keep changing, and how their sense of the system's fairness forms along the way.
I joined Meta. The work I have done since then is shown in The loop, above, where my Meta roles sit alongside the loop diagram.
I finished my Ph.D. from the Department of Computer Science at Purdue University.
I passed my early Ph.D. exam (called the preliminary exam).
I received the Graduate Teaching Award for my teaching in CS 57300.
Our paper on what people think is fair, and whether they stay was accepted to AIES 2022. Work done with Dr. Ming Yin.
Our paper on confirmation bias (favoring what you already think) in labels from many people was accepted to IJCAI 2021. Work done with Dr. Ming Yin.
I was awarded the Raymond Boyce Graduate Teaching Award for my teaching in CS 25000 in the Fall 2019 semester.
My department approved my request, and I officially became a PhD student.
I earned my bachelor's degree in Computer Science, with honors, from Purdue University.