Bohan Li
Tel: +86 15975558614
Email: milbh_brandon@mail.scut.edu.cn
Website: brandon-lee-bo.github.io
Education
South China University of Technology, Guangzhou, China
Sept. 2023 -- July 2027
B.Eng. in Microelectronics Science and Engineering
GPA: 3.81/4.00
Relevant Courses: Digital Electronic Circuits, Analog Electronic Circuits, Signals and Systems, Calculus, C++
Hong Kong University of Science and Technology
Aug. 2026 -- Present
Exchange Student, Hong Kong
Current coursework: Architecture for AI Systems, Computer Organization & Design, Generative AI.
Research Interests
Power & Performance Modeling, CPU Microarchitecture, Hardware-Software Co-design for AI.
Publications and Manuscripts
(† Equal contribution, * Corresponding author)
- Bohan Li†, Shengmin Li†, Xinyu Shi, Enyi Yao, Francky Catthoor, Simei Yang*. FlexVector: A SpMM Vector Accelerator with Flexible VRF for GCNs on Varying-Sparsity Graphs. IEEE Transactions on Computers (IF: 4.9, Accepted). Preprint: arXiv:2604.10113
- Bohan Li†, Xiangfeng Sun†, Ceyu Xu, Yuan Xie*. Not All Reuse Is Worth Keeping: Context-Aware Prefetch Retention for Shared Caches. HPCA 2027 (under rebuttal).
Research Experience
Shared-Cache Management: KEEPER
Research Intern, Hong Kong University of Science and Technology, supervised by Prof. Yuan Xie
Apr. 2026 -- July 2026
- Worked on KEEPER, a shared-LLC prefetch-retention policy combining per-fill reuse qualification with epoch-level source-core selection based on prefetch service and retirement blocking. Cross-core decisions remain off the LLC fill path.
- Implemented policies and baselines in ChampSim, ran 4-, 8-, and 16-core experiments, and assessed hardware feasibility through RTL synthesis. Drafted the manuscript and refined its argument with collaborators.
- Improved geometric-mean weighted speedup by 3.20% over Mockingjay and 1.56% over CARE across 80 heterogeneous four-core mixes, while preserving original priority for only 9.33% of admitted prefetch fills.
FlexVector: GCN Accelerator Design
South China University of Technology, supervised by Prof. Simei Yang
July 2024 -- Apr. 2026
- Designed a vector-based SpMM accelerator for GCN inference, using fixed and dynamic VRF regions to support irregular graph workloads. Combined inter-tile edge cutting and intra-tile vertex cutting with a hierarchical dataflow to reduce memory traffic.
- Led experiments and manuscript writing and revision. Used RTL and simulation to evaluate power, performance, and area, including the effects of graph preprocessing and storage organization.
- Evaluation on five real-world GCN datasets shows 3.78x speedup and 40.5% lower energy than a cache-centric baseline at comparable area and equal buffer capacity.
Frequency-Control Circuits for Microrobots
Summer Research Intern, Westlake University, supervised by Prof. Wei Wang
July -- Aug. 2025
- Learned analog IC design using Cadence Virtuoso. Designed and simulated a voltage-controlled ring oscillator, frequency dividers, and phase shifters for microrobot locomotion modes.
- Received the program's Best Poster Award for presenting the circuit design and its frequency, phase, power, and area tradeoffs.
Awards & Honors
- Best Poster Award, Westlake University Summer Research Program (2025)
- National First / Third Prize, College Student Chip and System Design Competition (2024 / 2025)
- National Third / South China First Prize, College IC Innovation Competition (2024)
- Second-Class Scholarship, South China University of Technology (2024, 2025)
Skills
- Programming and workflow: Verilog, Python, GitHub, LaTeX
- Architecture & circuit tools: ChampSim, CACTI, Synopsys Design Compiler, Vivado, Cadence Virtuoso
- Languages: Mandarin Chinese (native), English (IELTS 6.5)