2024-2025 Amazon ML Fellows
USC and Amazon have created a joint research center focused on development of new approaches to machine learning (ML) privacy, security, and trustworthiness. The Center for Secure and Trusted Machine Learning (in short, Trusted AI), which will be housed at the USC Viterbi School of Engineering, will support USC and Amazon researchers in the development of novel approaches to privacy-preserving ML solutions.
Zeyu Liu
Zeyu Liu
Biography
Zeyu Liu is a second-year Ph.D. candidate in Electrical Engineering at the University of Southern California, advised by Prof. Peter A. Beerel. As a member of the Energy Efficient Secure Sustainable Computing Group, his research centers on developing efficient and responsible deep learning systems.
Research Interests
During my PhD studies, I am committed to advancing efficient AI by leveraging state space models to develop fine-tuning methods that not only extend the capability of large language models (LLMs) to handle long contexts but also enhance privacy preservation. Additionally, I aim to integrate parameter-efficient fine-tuning (PEFT) techniques, such as LoRA, with safety alignment strategies to ensure that these methods are inherently safety-oriented.
What does being an Amazon ML Fellow mean to you?
“Being named an Amazon ML Fellow is a tremendous honor, recognizing my previous work and achievements while fueling my pursuit of impactful research. With the fellowship’s support, I am better equipped to delve deeper into my ideas and make my contributions to both the field and society.”
Huihan Li
Huihan Li
Biography
Huihan Li is currently a third-year PhD student working on Natural Language Processing at the University of Southern California, advised by Prof. Xiang Ren. Her research on trustworthy natural language processing systems focuses on training and evaluating language models for better generalization, as well as next-generation data selection for model improvement. Previously, she got her M.S.E in Computer Science from Princeton University, advised by Prof. Danqi Chen.
Research Interests
My work focuses on testing and achieving generalization through the long-tail of model distribution, where models are more prone to undesirable behaviors over the distribution or domain in which it has low confidence. My broad research goal of generalization in long-tail distribution comprises two milestones: (1) efficient, general algorithms for long-tail data detection and generation in diverse domains (2) developing model training and alignment strategies that guarantee generalization in the long-tail scenarios.
What does being an Amazon ML Fellow mean to you?
“I am deeply honored to be recognized as an Amazon ML Fellow. This fellowship provides me with the opportunity to explore innovations in long-tail generalization and contribute to the journey of trustworthy AI in shaping the field.”
Tejas Srinivasan
Tejas Srinivasan
Biography
Tejas is a fourth-year PhD student in Computer Science, advised by Prof. Jesse Thomason as part of the GLAMOR Lab. He has completed internships at the Allen Institute for AI and Microsoft Research, and previously worked as a Research Scientist at the AI Foundation. Outside of work, Tejas spends most of his time playing mediocre tennis and re-watching episodes of Community.
Research Interests
My research interests lie in user-centric approaches to building reliable AI systems, particularly Large Language Models. While the Machine Learning community has developed methods for estimating and calibrating model confidence, not much consideration has been given to whether they improve reliability for the user. I plan to study how methods for computing and expressing model uncertainty can best be integrated into AI systems to foster reliable AI use by human users
What does being an Amazon ML Fellow mean to you?
What it means to be an Amazon Fellow: It is a great honor to be selected as an Amazon ML Fellow from a pool of talented PhD students who are all working on important problems. Further, it is incredibly gratifying to see an industry giant like Amazon value user-focused AI research, especially when the arms race to develop the best LLM can sideline human-centric approaches that require engaging with and understanding the needs of users. It is heartening that Amazon recognizes the importance of research that examines how humans use and trust AI systems.