Ruixuan Liu

Ruixuan Liu is a Postdoctoral Associate in the Department of Biomedical Informatics & Data Science at Yale School of Medicine, working with Prof. Lucila Ohno-Machado. Her research advances trustworthy, privacy-preserving machine learning, with a focus on differential privacy, federated learning, and the privacy and security of large language models and foundation models. She investigates how AI systems expose sensitive information and develops defenses that combine rigorous privacy guarantees with computational efficiency and practical deployment, including applications to health and genomic data.

Her recent work establishes methods for measuring data extraction risk in LLM APIs (IEEE S&P 2026), protects web content against unauthorized crawling and LLM training (NDSS 2026), and reveals how pre-trained language models can compromise the privacy of fine-tuning data (ACM CCS 2024). She also develops differentially private optimization methods that reduce or eliminate hyperparameter tuning (ICLR 2025 Spotlight), building on earlier research in privacy-preserving federated learning published at AAAI and ACM SIGKDD. Her co-authored work received an Outstanding Paper Award at ACM MM 2025. Her methods for differentially private training have been incorporated into Amazon AWS AI Lab's open-source FastDP library and built upon in Dynamo AI's DynamoEnhance product.

Before joining Yale, she was a Postdoctoral Fellow in the AIMS Lab at Emory University, working with Prof. Li Xiong, and co-taught the graduate course Data Privacy and Security. She earned her Ph.D. in Computer Science from Renmin University of China, advised by Prof. Hong Chen, and held research internships at Amazon AWS AI Lab and Microsoft Research Asia.

She serves on the program committees of ACM CCS, IEEE SaTML, and AAAI, and was recognized as an ACM SIGKDD Commendable Reviewer in 2025.