# Yuyi Li

CS PhD Student University of California, Davis [CV](https://drive.google.com/file/d/1K9UAr3A8Co_C5JIKV4dH3GYVHgvorLjQ/view?usp=sharing) [Email](mailto:nylli@ucdavis.edu) [GitHub](https://github.com/Yuy1L1) [LinkedIn](https://www.linkedin.com/in/vivian-yuyi-li-16b78a210/) [X](https://x.com/YuyiLi5) # About Me Hi, my name is Yuyi. I’m a second-year PhD student at [UC Davis](https://cs.ucdavis.edu/graduate), where I’m fortunate to be advised by [Amanda Raybuck](https://web.cs.ucdavis.edu/~araybuck/). Before joining the PhD program, I worked with Prof. Jason Lowe-Power on computer architecture and memory systems, which sparked my interest in OS-level performance. I did my undergrad at Davis as well (yes, I never left). I majored in Computer Science with a Math minor, mostly because I liked proving things wrong using counterexamples. My research interests include ML Systems, Networked Systems, and Memory Systems. In summer 2026, I was a Research Intern at [Microsoft Research](https://www.microsoft.com/en-us/research/group/systems-research-group-redmond/), Redmond. I worked with [Dr. Dan R. K. Ports](https://drkp.net/) and Dr. Jacob Nelson on scale-out network in PD Disaggregated LLM Serving systems. # Current Projects **Optical Networking for Disaggregated LLM Serving** We want to explore whether constrained optical switching can replace existing switching in cross-node PD disaggregated LLM serving system without significant performance degradation. **Agentic Memory Manager for Tiered Memory systems** We want to answer if an agent can use memory-access traces to predict future accesses and make better paging decisions in DRAM+CXL tiered-memory systems. # Past Projects **LLM for GPU Communications** We benchmarked and verified that LLM-generated GPU communication code needs systematic evaluation across hardware and communication stacks before it can be trusted. [Blog](https://uccl-project.github.io/posts/commbench/) [Paper](https://arxiv.org/pdf/2608.04450) [Code](https://github.com/uccl-project/CommBench/tree/main) **FairMem (QoS-Aware Tiered Memory)** We improved allocation fairness of two-tiered memory system in a multi-tenant settings. **gem5 simulator GPU Library** We created a modular GPU frontend library for gem5, enabling high-fidelity and reproducible simulation of GPU workloads. # Teaching - ECS 122A: Undergraduate Algorithm Design. Teaching Assistant. Winter 2026 - ECS 120: Undergraduate Theory of Computation. Teaching Assistant. Fall 2023 - [ECS 189G](https://github.com/ucdavis/FairMLCourse): Ethics in Machine Learning. Teaching Assistant. Spring 2023 # Service - Artifact Evaluation Committee: [SOSP '26](https://sysartifacts.github.io/sosp2026/committee), [EuroSys '26](https://sysartifacts.github.io/eurosys2026/organizers), [Sigcomm '25](https://conferences.sigcomm.org/sigcomm/2025/cf-artifacts/#:~:text=Organizers) - Program Committee: [AgentSkills '26](https://www.agentskills-workshop.org/), [AISec '26, '25](https://aisec.cc/) - Shadow PC: [Sigmetrics '26](https://www.sigmetrics.org/sigmetrics2026/shadowpc.html) # Awards - ACM Sigcomm Student Travel Grant 2026 - ASPLOS Student Travel Grant 2026 - OSDI Student Travel Grant 2025 # Talk(Not selected) - [NorCal Systems Day](https://norcalsystemsday.gitlab.io/program.html) # Miscellaneous and Blog Posts Apart from research, I split my time between biking and playing questionable notes on my electric bass when I think no one’s listening. It’s jazz. Probably;) I occasionally open [Project Euler](https://projecteuler.net/) just to scroll past problems I don’t understand. If you're into math elegance (or masochism), here's [21 proofs of Euler’s formula](https://ics.uci.edu/~eppstein/junkyard/euler/) that I find interesting. I also (try to) keep a [blog](blog.html) where I write about CUDA learning notes. Feel free to reach out if you like systems, jazz, or jazzy systems.