Email | LinkedIn | Google Scholar

Wanyu Zhao

I am a second-year PhD at UIUC CS, advised by Prof. Brighten Godfrey. I’ve also had the privilege of working with Prof. Charith Mendis.

My research interests lie in computer systems, machine learning, and the intersection between the two areas. I’ve built practical and efficient systems for machine learning (Sys4ML). My recent research focuses on applying machine learning to systems design (ML4Sys).

I’m excited about opportunities where ML not only outperforms traditional heuristics but also unlocks system designs once unimaginable. From advancing ML techniques–such as explainability and efficiency–to designing next-generation system architecture integrating ML, I hope ML advancements will drive continuous progress in computer systems, much like Moore’s Law fueled decades of progress.

I’m part of LDOS, where we are building the next-generation Machine Learning-based Operating System to drive computing infrastructure toward high efficiency and performance.

News

• [Oct. 2024] I gave a talk on optimizing TGNN training and inference at UIUC compiler seminar. [video]
• [Oct. 2024] I co-hosted a tutorial session on efficient temporal graph learning at CIKM ‘24. [tutorial website]
• [Mar. 2024] I gave a talk on optimizing TGNN training and inference in TGL Group. [video]
• [Apr. 2024] We released the TGLite framework for programming temporal GNNs. [code] [documentation]
• [Aug. 2023] Join UIUC!

Research

• Efficient Temporal Graph Learning: with an emphasis on efficient and scalable temporal graph neural network (TGNN) training and the development of user-friendly frameworks for temporal graph learning.

  • [In submission] TGCache: Reducing Data Movement for Large Scale Temporal GNN Training

    Zhao, W., Ge, Y., Sudusinghe, C., Mendis, C.

  • [CIKM ‘24] Towards Efficient Temporal Graph Learning: Algorithms, Frameworks, and Tools

    Wang, R., Zhao, W., Sun, D., Mendis, C., Abdelzaher, T.

  • [SIGIR ‘24] TGOnline: Enhancing Temporal Graph Learning with Adaptive Online Meta-learning

    Wang, R., Huang, J., Zhang, Y., Li, J., Wang, Y., Zhao, W., Liu, S., Mendis, C., Abdelzaher, T.

Misc

I’ve been lucky to work with Prof. Jian Huang on learning-based software-defined flash infrastructure at UIUC and Prof. Anastasios Kyrillidis on distributed matrix completion at Rice University during my undergrad.

I like coding and I’m committed to open source. Some of my past projects include: ExAllTrue (won the second place in the financial track of DeeCamp 2021), Rapq, MazeGenerator, and ClassChat.

About Me

• I love music. I am an amateur dancer (Hiphop, Chinese folk dance, Latin, Ballet). I also play some Chinese traditional musical instruments.

• My twin sister works on quantum information theory.