Jiajun Qin
I am Jiajun Qin (秦嘉俊 in Chinese), a second-year PhD student at the Department of Computer Science and Engineering, The Chinese University of Hong Kong (CUHK) since 2025 Fall, advised by Prof.Bei Yu. Prior to that, I graduated from Zhejiang University (Turing Honors Class) in 2025.
I am open to research collaborations and internship opportunities. If you are interested in working together or discussing potential projects, feel free to contact me at hobbitqia@gmail.com!
Research Overview
My research is on hardware/software co-design for efficient and adaptable AI acceleration. An AI accelerator must be specialized enough to be efficient and adaptable enough to follow rapidly evolving models, and meeting both demands requires decisions that cut across the stack rather than within a single layer. My work therefore targets three interfaces of the stack: between algorithms and the hardware that executes them, between the compiler and a reconfigurable architecture, and between the accelerator and the system around it.
1. Algorithms and hardware for emerging AI workloads.
I co-design approximations and hardware for LLM nonlinear operators, whose growing inference cost and rapid evolution demand both efficiency and adaptability.
- HAAN (DATE 2025) accelerates normalization by predicting statistics across adjacent layers instead of recomputing them, with hardware built around this prediction.
- PICACHU (ASPLOS 2025) generalizes this to all nonlinear operators in LLMs by combining algorithms for their shared primitives with a heterogeneous CGRA and a compiler that maps any such operator onto it.
2. Compilers and architectures for reconfigurable fabrics.
I develop compilers and reconfigurable architectures together, pairing mapping techniques with hardware support and automating workload-driven hardware customization.
- FLAME (DATE 2026) treats operation latency as a mapping constraint and pairs three execution strategies for multi-cycle operations with the hardware support each requires.
- NUKE (ICCAD 2026) automates pattern fusion: with e-graphs, algebraic rewrites and fusion decisions are explored jointly, so the hardware patterns a CGRA should provide are derived from the workload rather than extracted by hand.
- Aether (HPCA 2027) customizes CGRAs through coordinated kernel-level and pattern-level fusion, co-designing these compiler optimizations with hardware support for multi-cycle execution.
3. Accelerators in complete systems (ongoing).
I integrate reconfigurable fabrics into heterogeneous SoCs and develop programming models and compilers for whole-model megakernels, extending kernel-level gains to complete applications.
News
I will attend ICCAD 2026 in San Jose (Nov 9–11) and ICCD 2026 Student Research Forum in Hong Kong (Nov 16–18). Feel free to reach out if you’d like to meet and discuss!
- Nov/2026: Our paper Aether has been accepted to HPCA 2027!
- Jul/2026: Our paper NUKE has been accepted to ICCAD 2026!
- Apr/2026: I presented FLAME at DATE 2026 in Verona, Italy (Apr 20–22). ().
- Nov/2025: Our papers FLAME and MAEDA have been accepted to DATE 2026!
- Jun/2025: I graduated from Zhejiang University, completing my B.Eng. in Computer Science and Technology in the Turing Honors Class ().
- Apr/2025: I presented PICACHU at ASPLOS 2025 in Rotterdam, the Netherlands (Mar 30–Apr 3) (video).
- Feb/2025: Our paper PICACHU has been accepted to ASPLOS 2025!
Selected Publications
For a completed list of publications, see this.
- ASPLOS’25 PICACHU: Plug-In CGRA Handling Upcoming Nonlinear Operations in LLMs. Jiajun Qin, Tianhua Xia, Cheng Tan, Jeff Zhang, Sai Qian Zhang. ACM International Conference on Architectural Support for Programming Languages and Operating Systems, 2025.
- DATE’26 FLAME: A Framework Exploring Execution Strategies for Multi-Cycle Operations in CGRA. Jiajun Qin, Cheng Tan, Ruihong Yin, Tianhua Xia, Sai Qian Zhang, Bei Yu. IEEE/ACM Proceedings Design, Automation and Test in Europe, 2026.
- ICCAD’26 NUKE: An Automatic Pattern Fusion Framework for CGRAs using E-graphs. Jiajun Qin, Yuhao Ji, Zhaotan Lin, Cheng Tan, Xinyun Zhang, Bei Yu. IEEE/ACM International Conference on Computer-Aided Design, 2026.
- HPCA’27 Aether: An Agile Framework for CGRA Customization via Coordinated Hierarchical Fusion. Jiajun Qin, Zhaotan Lin, Shangkun Li, Cheng Tan, Wei Zhang, Bei Yu. IEEE International Symposium on High-Performance Computer Architecture, 2027.
Education
- The Chinese University of Hong Kong
- Ph.D. in Computer Science and Engineering, 2025.8 - Now
- Zhejiang University
- B.Eng. in Computer Science and Technology, 2021.9 - 2025.6
- Turing Class, Chu Kochen Honors College
Experiences
- China Mobile (Hong Kong) Innovation Research Institute, Hong Kong SAR, China
- PhD Intern, 2026.04 - 2026.08
- Topic: AI-assisted Development
- ChatEDA Technology Co., Ltd, Shenzhen, China
- Research Intern, 2025.05 - 2026.03
- Topic: LLM for EDA
- SAI Lab, New York University
- Remote Research Intern, 2024.3 - 2024.10
- Supervisor: Prof. Sai Qian Zhang
- Collaborator: Dr. Cheng Tan and Prof. Jeff Zhang
- Topic: SW/HW Co-Design and Reconfigurable Accelerator for LLMs
- RC4ML, Zhejiang University
- Research Intern, 2023.8 - 2024.9
- Supervisor: Prof. Jie Zhang and Prof. Zeke Wang
- Topic: FPGA-based SmartNIC
Academic Services
- Journal Reviewer: IEEE TCAD
- Artifact Evaluation Committee: ASPLOS 2027, MICRO 2026
- External Reviewer: FPL 2026
- Secondary Reviewer: EuroSys 2027, HPCA 2027, ASP-DAC 2026, ICCAD 2026, NeurIPS 2025, DAC 2025
Selected Honors & Awards
- Zhejiang Government Scholarship, 2022, 2023, 2024
- Huawei Elite Scholarship, 2023
- Third Prize in National Student Computer System Capability Challenge, 2023
- Second Prize in CSP-S 2019
Teaching Assistants
At The Chinese University of Hong KOng:
- ENGG1110 - Problem Solving By Programming with Dr.Calvin Kam, 2026 Fall
At Zhejiang University:
- CS1029M - Introduction to Computing Systems with Prof.Yale Patt and Prof.Xiaohong Jiang, 2023 Summer
- CS2056M - Computer Organization and Design, with Prof.Haifeng Liu, 2024 Spring & Summer
- CS1028M - Computer SystemⅠ, with Prof.Li Lu and Prof.Haoting Shen, 2024 Spring & Summer
- CS2051M - Computer Architecture, with Prof.Rui Chang, 2024 Fall & Winter
- CS3125M - Assembly Language And Microcomputer Interface, with Prof.Ming Cai, 2024 Fall & Winter
- CS1001G - Lab. & Fundamentals of C Programming, 2024 Fall & Winter
- CS1028M - Computer System Ⅰ, with Prof.Haifeng Liu, 2025 Spring & Summer
- CS1019F - Digital Logic and Design, with Prof.Ming Cai, 2025 Spring & Summer
Skills
- Programming Languages: C/C++, Verilog, Chisel, Python, Rust
- Tools: Pytorch, LLVM, Vivado, LaTeX, Pymtl
