cv
General Information
| Name | Zhongming Yu |
| zhy025@ucsd.edu | |
| Location | San Diego, California |
| Languages | English, Mandarin |
Education
- 2022 - present
Ph.D. in Computer Science and Engineering
University of California, San Diego
- Advised by Prof. Jishen Zhao.
- Research: LLM agent infrastructure, agent memory, and ML systems.
- 2018 - 2022
B.E. in Electronic Engineering
Tsinghua University
- Sparse computation and GPU acceleration for graph learning.
Experience
- Jun. 2026 - present
Deep Learning Compiler Intern
NVIDIA
- 2025
Student Researcher
Google
- Evolving LLM agent frameworks for machine learning.
Open Source
- 2026 - now
CodeNib
- A multi-view data system serving repository context to coding agents.
- 2025 - now
OrcaLoca
- An LLM agent framework for software issue localization (ICML '25).
- 2024 - now
GeoT
- Tensor-centric GNN library built on efficient GPU segment reduction.
- 2021 - 2023
dgSPARSE
- GPU sparse kernels and GNN operators for graph learning.
Service
- 2023 - 2024
- Program Committee, ACM SIGKDD (KDD '23, '24, '25)
- Poster Reviewer, ES-FoMo-II Workshop @ ICML '24
Honors and Awards
Research Interests
-
Agent infrastructure
- Repository context serving and incremental code indexes.
- Retrieval that respects code structure rather than flat text.
-
Agent memory
- Multi-agent memory hierarchies and consistency protocols.
- Long-horizon retention across states, actions, and tool outputs.
-
Evaluation
- Benchmarks for continuous software evolution.
-
ML systems
- Sparse kernels, sparse attention, and GPU compilation.