Incoming CS PhD student at Princeton (Fall 2026), advised by Tri Dao and Ravi Netravali. Previously a visiting student researcher at UC Berkeley's Sky Computing Lab with Ion Stoica. B.S. from Renmin University of China.

I work broadly on machine learning systems, with interests in efficient infrastructure, training and inference systems, and the design of systems for emerging AI workloads.

• SkyNomad: On Using Multi-Region Spot Instances to Minimize AI Batch Job Cost. NSDI '27.
• LEANN: A Low-Storage Vector Index for Personal Devices. MLSys '26, Best Paper Award.
• FrontierCS: Evolving Challenges for Evolving Intelligence. ICML '26.
• PixelRAG: Web Screenshots Beat Text for Retrieval-Augmented Generation. Preprint.

• LEANN, 12k stars: a vector index that cuts storage by 97%, so RAG runs locally on a laptop.
• PixelRAG, 9k stars: retrieval over web screenshots instead of parsed text.
• SkyPilot, 10k stars: top-10 contributor; built the high-availability controller for SkyServe.

zhifei.li · Google Scholar · andylizf@gmail.com