Hojae Son

pic.jpg

Hi, I'm Hojae Son, a Ph.D. candidate in Computer Science at UMass Amherst, advised by Professor Marco Serafini. My research interests are in systems for machine learning, especially LLM inference, KV cache optimization, vector databases, and graph neural network systems. I am also a Research Assistant for the Unity Cluster at the UMass Amherst Center for Data Science and Artificial Intelligence.

Previously, I worked as a Research Engineer at Hyundai Mobis and lululab (Samsung C-Lab).

news

Apr, 2026 Our paper, “PRISM: A Training System to Unlock the Potential of Temporal Graph Learning Through Staleness Avoidance,” has been accepted to VLDB 2026.
Dec, 2024 Our paper, “Graph Neural Network Training Systems: A Performance Comparison of Full-Graph and Mini-Batch,” has been accepted to VLDB 2025.

selected publications

  1. arXiv
    Kalypso: Relational LLM Serving
    Hojae Son, Md Ashraful Islam, Huy Gia Cao, and 2 more authors
    2026
  2. PVLDB
    PRISM: A Training System to Unlock the Potential of Temporal Graph Learning Through Staleness Avoidance
    Md Ashraful Islam, Hojae Son, Suhaas Kiran, and 1 more author
    In PVLDB, 2026
  3. PVLDB
    Graph Neural Network Training Systems: A Performance Comparison of Full-Graph and Mini-Batch
    Saurabh Bajaj, Hojae Son, Juelin Liu, and 2 more authors
    In PVLDB, 2025