CV
General Information
| Full Name | Yifei Yang (杨翼飞) |
| Affiliation | School of Computer Science, Sun Yat-sen University |
| Current Position | Master Student |
| contact@yfyang.me | |
| Research Interests | High-Performance Computing Systems, Heterogeneous Computing, Scheduling Optimization |
Education
-
2024.09 - 2027.06
(Expected)Master of Engineering in Computer Technology
School of Computer Science, Sun Yat-sen University - Research focus on high-performance computing systems and heterogeneous scheduling optimization
-
2020.09 - 2024.06 Bachelor of Engineering in Software Engineering
School of Software Engineering, Sun Yat-sen University - GPA: 3.9/4.0
- Received recommended admission to postgraduate study without examination
Selected Projects
-
Multi-Application Scheduling Framework for Heterogeneous Supercomputing Systems
- National Key R&D Program of China, key High-Performance Computing project under the 14th Five-Year Plan
- Period: 2023.12 - Present
- Built a supporting environment and development framework for next-generation domestic heterogeneous supercomputers, covering distributed scheduling, heterogeneous runtime adaptation, and AI / scientific computing workloads.
- Ported GASNet-EX and UCX on a next-generation Sunway platform by hooking the MPI backend and exposing an in-house RDMA interface, delivering 10%+ better network performance than the conventional MPI path in target scenarios.
- Improved UCX environment support on a next-generation Tianhe supercomputer and achieved 10+ GB/s inter-node bandwidth utilization between neighboring nodes.
- Ported Stanford Legion and adapted the operator middleware required by FlexFlow and Lux on two domestic supercomputing platforms.
- As a core member of the lead institution, contributed to project coordination, reporting, third-party evaluation, and on-site inspection support.
-
Content Safety Firewall for LLM Inference
- Enterprise collaboration project (commercialized)
- Period: 2025.09 - Present
- Built a content-safety firewall for LLM inference services with real-time risk detection, response interception, and safety evaluation / ranking for open-source models on domestic heterogeneous hardware.
- Independently deployed the risk-detection model on Ascend NPUs and T-Head PPUs without vendor technical support, achieving 2x+ single-card inference acceleration on Ascend 910B1 for production workloads.
- Participated in system architecture design, technology selection, infrastructure setup, deployment encryption design, and automation for full delivery from prototype to production.
-
NuttX RTOS SMP Scheduler Optimization
- National Computer System Development Capability Competition project
- Period: 2023.03 - 2023.08
- Designed and implemented a POSIX-compatible per-core scheduler on Apache NuttX RTOS to remove the global ready-queue bottleneck in SMP environments.
- Validated the solution on ARM and RISC-V platforms in both simulation and real hardware, and assisted in diagnosing NuttX native Arm soft-interrupt response issues.
- Achievement: Second Prize in the National Finals
- [Slides]
Selected Highlights
- Published 3 English conference papers as first author or graduate first author, including 1 Best Paper Award.
- Granted 1 invention patent and registered 1 software copyright as student first author.
- Achieved National Second Prize in the National Computer System Development Capability Competition (OS track), MCM Finalist, and top 2.26% nationwide in the 30th CCF CSP C/C++ group.
- Recipient of the Sun Yat-sen University–MUCFC Scholarship, the First Prize of the Sun Yat-sen University Scholarship for Outstanding Students, and the Sun Yat-sen University First-Class Graduate Grant.
Publications
- Yifei Yang, Tianyufei Zhou, Linchang Xiao, Chengrun Yang, Xuezheng Liu, Miao Hu, Di Wu. "NAAM: Enhancing Automatic Task Mapping Efficiency on NUMA Machines" In Parallel and Distributed Computing, Applications and Technologies (PDCAT 2025).
- Yifei Yang, Linchang Xiao, Tianyufei Zhou, Chengrun Yang, Xuezheng Liu and Miao Hu. "DRL-MOSHRS: A Deep Reinforcement Learning Approach for Multi-Objective Scheduling in Heterogeneous HPC Systems" In the 11th IEEE International Conference on High Performance and Smart Computing (HPSC 2025).
- Tianyufei Zhou, Yifei Yang, Chengrun Yang, Linchang Xiao, Xuezheng Liu and Miao Hu. "TaskFlare: A Heterogeneity-aware Unified Scheduling Framework for Diverse Domain-Specific Applications in Supercomputing Environments" In the 11th IEEE International Conference on High Performance and Smart Computing (HPSC 2025). [Best Paper Award]
Technical Skills
-
Programming Languages
- Proficient: C/C++
- Familiar: Rust, Python
- Working knowledge: Go, TypeScript
-
Systems and Tools
- Linux system administration and performance tuning
- Git, Docker, CMake, xmake
- GDB, Perf, eBPF
- CUDA, ROCm
-
Technical Areas
- Heterogeneous computing and performance optimization
- High-performance computing and parallel programming
- Operating system kernel development
- Domestic hardware and operating system ecosystems
-
Framework Experience
- Mainstream supercomputing scheduling software
- Stanford Legion runtime
- TVM and LLVM compiler toolchains
- Apache NuttX RTOS
Open Source Contributions
-
VSCode
- Fixed the GLibcxx version detection bug for VSCode 1.85.0 on Linux and resolved the related shell-script vulnerability; the patch was merged upstream.
-
Parallel Programming Course Tooling
- Designed and open-sourced course experiment frameworks and several practical utilities for parallel programming classes and lab environments.