Enable LLM Internal Observability
Treating model-internal observability as a first-class systems primitive for high-performance LLM inference.
•
1 min read
About Me
I am a third-year Ph.D. student in the Department of Computer Science at the University of Maryland, College Park, advised by Prof. Zaoxing Liu. Before joining UMD, I received my B.S. in Computer Engineering from Boston University.
My research primarily centers on two directions: (1) ML systems, from ML-for-systems to systems-for-ML, with a current focus on telemetry and observability for LLM platforms; (2) Data-driven machine learning, where I develop algorithms for learning from complex, real-world data.
Education
Ph.D. in Computer Science
2023 – Expected 2028
University of Maryland
B.S. in Computer Engineering
2019 – 2023
Boston University
Interests
Large Language Model
Machine Learning Systems
Distributed Systems
Selected Projects view all →
Collective Communication Under Failures for Distributed ML
Closing the gap between fail-stop CCLs and the realities of large-scale GPU training/serving — through resilient communication and bandwidth-optimal AllReduce under failures.
•
1 min read