Hi, I'm Donna.
I'm a Rackham Scholar and a first-year PhD student in the Computer Science & Engineering Division at the University of Michigan, Ann Arbor. I'm fortunate to be advised by Prof. Lin Ma, with whom I've worked since my undergraduate years. That is why I chose to continue my doctoral studies under his guidance.
My research lies at the intersection of database management systems and machine learning. Currently, I focus on optimizing index structures for large language models, including hierarchical indexing for faster, more scalable retrieval.
I worked across industry, most recently at Google on modeling infrastructure, and earlier at Citadel, Tesla, NVIDIA, and Ford. I'm originally from the Bay Area, California.
After my PhD, I plan to work in industry on database research, mainly indexing. I want to keep developing new hierarchical indexing methods that help LLMs work with large collections of text.
Research
Hierarchical document indexing for LLM-based analysis
LLMs have trouble with tasks that need information spread across a long document or a large set of documents. Most retrieval pipelines cut text into flat chunks and lose the structure of the original document. In this project I build a hierarchical plain-text index inspired by the B+ tree, so an LLM can start from a high-level view of a document and work down to the passages it needs. We evaluate it on the NarrativeQA and CUAD benchmarks. An early version won Best Paper at the VLDB 2026 Ph.D. Workshop, and the full paper, Rethinking Document Indexing for LLM-based Analysis, will appear at CIDR 2027.
LLM-based query plan representation for cost and cardinality estimation
A query optimizer picks a plan based on its estimates of cost and cardinality, and those estimates are often wrong. This project looks at how to represent query plans so that an LLM can use them to predict cost and cardinality more accurately. We evaluate the approach on the TPC-H, TPC-DS, IMDB, and STATS workloads.
Publications
Donna Pham, Zhongwei Xu, Siyuan Dong, Haotian Gong, Lin Ma
17th Conference on Innovative Data Systems Research (CIDR 2027), Amsterdam PDF
Donna Pham
VLDB 2026 Ph.D. WorkshopBest Paper arXivPoster
Zhongwei Xu, Siyuan Dong, Haotian Gong, Donna Pham, Lin Ma
Proceedings of the VLDB Endowment (PVLDB), Vol. 19, No. 8, 2026 DOIPDF
Experience
Contact
Email: phdonn@umich.edu
Computer Science and Engineering, University of Michigan, Ann Arbor, MI
LinkedIn