Zhifei Yang

Muse Infrastructure @ Meta

I am a systems engineer and researcher working on Muse, Meta's personal AI assistant. I build the VM infrastructure behind it, with a focus on performance optimization. Previously, I worked on the performance and reliability engineering of CPU/GPU system software for large-scale AI workloads at TikTok/ByteDance.

I received a master's degree in Computer Science from ETH Zürich.

Publications & Talks

1. Boosting Network Performance of Confidential VM Using Userspace Stack
   William Lam, Yuwei Zhang, Zhifei Yang
   In DPDK Summit 2024 (video).
2. Fast and Secure: DPDK Meets Confidential Computing
   Zhifei Yang, Liang Ma
   In DPDK Summit 2023 (video).
3. Odinfs: Scaling PM performance with Opportunistic Delegation
   Diyu Zhou, Yuchen Qian, Vishal Gupta, Zhifei Yang, Changwoo Min, and Sanidhya Kashyap
   In Proceedings of the 16th USENIX Symposium on Operating Systems Design and Implementation (OSDI '22).
4. Motivating High Performance Serverless Workloads
   Hai Duc Nguyen, Zhifei Yang, Andrew A. Chien
   In Proceedings of the 1st Workshop on High Performance Serverless Computing (HiPS '21).

Experiences

• Production Engineer, Meta.
  London, UK, 11.2025 ~ Now.
  Helping scale Meta's VM infrastructure for secure AI-agent runtime and frontier model training to millions and billions of users.
• System Software Engineer, TikTok/ByteDance.
  System Technologies & Engineering - London, UK, 2023 ~ 2025.
  Performance tuning and reliability engineering of TikTok/ByteDance's global system software, focusing on high-performance GPU infrastructure and the Linux kernel.
• Research Assistant, School of Computer and Communication Sciences, EPFL.
  Supervisors: Prof. Sanidhya Kashyap and Prof. Babak Falsafi. 09.2021 ~ 12.2022.
  Studied ways to achieve near-zero virtualization overhead and scalable virtual memory address translation
• Research Assistant, Large-scale Sustainable Systems Group, The University of Chicago.
  Advisor: Prof. Andrew A. Chien. 09.2020 ~ 06.2021.
  Contributed to Realtime Serverless, a novel cloud software architecture to enhance Function-as-a-Service with application-defined performance guarantee
• Research Assistant and Master Thesis, Systems Group, ETH Zurich.
  Advisor: Prof. Gustavo Alonso. 2018 ~ 2020.
  Developed various database operations on FPGA for better performance and energy efficiency
• Research Intern, Systems Research Group, Microsoft Research Asia.
  Mentors: Dr. Hucheng Zhou and Dr. Lintao Zhang. 06.2016 ~ 04.2017.
  Optimized the distributed training performance of a popular ML model: gradient boosting decision tree

Teaching

• TA of CS-173 Digital System Design, EPFL. Spring 2022.
  
• TA of High-level Language Programming, Harbin Institute of Technology. 2014 ~ 2015.
  

Services

• OSDI'22 and ATC'22 Artifact Evaluation Committee. 2022.
  
• PPoPP'22 Artifact Evaluation Committee. 2021.
  

To seek, to grind, to create.