Kaiyan Chang (常开颜)
Institute of Computing Technology, Chinese Academy of Sciences
State key lab of Processor Chip
changkaiyan [AT] live [DOT] com
Hi! I received my Ph.D. degree in Computer Architecture from the Institute of Computing Technology, Chinese Academy of Sciences in June 2026, advised by Prof. Ying Wang and Prof. Huawei Li. Before that, I entered the undergraduate program in Software Engineering at the University of Electronic Science and Technology of China in 2017 and received my B.Eng. degree in 2021, advised by Prof. Wei Jiang.
My research lies at the intersection of computer architecture, electronic design automation, and large language models. I am particularly interested in building intelligent systems that can understand, generate, verify, and optimize hardware designs. My recent work includes ChipGPT, ChipGPT-FT, multimodal Verilog generation, and LLM-based cost modeling for dataflow accelerators.
During my Ph.D., I published multiple first-author or co-first-author papers at top computer architecture and EDA venues, including MICRO, DAC, ICCAD, TACO, and SCIS. My work spans natural-language-driven chip design, LLM performance modeling, chip design data augmentation, and intelligent agent workflows. My most cited first-author paper has received more than 200 citations; related work was selected among the Trending CS Papers Top 8 in computer architecture and covered by outlets including China Electronics News. I have received honors including the President’s Special Scholarship of the Chinese Academy of Sciences, the highest honor granted by CAS to graduate students; the Director’s Special Award of the Institute of Computing Technology, CAS; and DAC Young Fellow.
您好!我于 2026 年 6 月在中国科学院计算技术研究所获得计算机体系结构方向博士学位,导师为王颖研究员和李华伟研究员。在此之前,我于 2017 年进入电子科技大学软件工程专业学习,并于 2021 年获得工学学士学位,本科导师为江维教授。
我的研究方向位于计算机体系结构、电子设计自动化和大语言模型的交叉领域。我尤其关注如何构建能够理解、生成、验证和优化硬件设计的智能系统。近期工作包括 ChipGPT、ChipGPT-FT、多模态 Verilog 生成,以及面向数据流加速器的基于大语言模型的代价建模。
博士期间,我以第一作者或共同第一作者在 MICRO、DAC、ICCAD、TACO、SCIS 等计算机体系结构与电子设计自动化领域顶级会议和期刊发表多篇论文,研究成果涵盖自然语言芯片设计、LLM 性能建模、芯片设计数据增强与智能 Agent 工作流等方向。一作单篇最高被引 200 余次,相关成果曾入选计算机体系结构 Trending CS Papers Top 8,并受到《中国电子报》等媒体报道。曾获中国科学院院长特别奖(中国科学院授予研究生的最高荣誉)、中国科学院计算技术研究所所长特别奖、DAC Young Fellow 等荣誉。
Background and internships 教育背景与实习经历
Before and during my Ph.D., I interned at Intel Asia-Pacific R&D Center, Huawei 2012 Laboratories, and the National Innovation Center for EDA, where I worked on compiler testing and development, AI compiler research, and LLM-based circuit code generation.
博士期间及之前,我曾在英特尔亚太研发中心、华为 2012 实验室、国家集成电路设计自动化技术创新中心实习,从事编译器测试与开发、AI 编译器研究以及基于大语言模型的电路代码生成等工作。
ChipGPT for Processor Design Automation 面向处理器设计自动化的 ChipGPT
My research explores how large language models can understand architecture intent, generate and refine hardware code, and support processor design workflows with domain knowledge and tool feedback.
我的研究探索大语言模型如何理解体系结构设计意图、生成与优化硬件代码,并结合领域知识和工具反馈支撑处理器设计流程。
- Natural-language to processor design intent
- Knowledge-grounded Verilog and microarchitecture generation
- Evaluation, verification, and design-flow integration
- 从自然语言理解处理器设计意图
- 基于知识增强的 Verilog 与微体系结构生成
- 面向评估、验证和设计流程集成的方法
News 动态
- My Ph.D. dissertation defense was successfully completed.
- Our LLMulator paper was accepted to MICRO 2025.
- ChipGPT-related research was selected as a Top 8 trending paper in computer architecture by Trending CS Papers.
- Presented ChipGPT, a tool for AI-assisted chip design, at the 2024 Public Science Day frontier technology showcase of the Chinese Academy of Sciences.
Selected publications 代表性论文 All publications 全部论文
Research 研究方向
01 Natural-Language Processor Design Automation 自然语言驱动的处理器设计自动化 Bridging architecture intent, design knowledge, and automated hardware generation 连接体系结构设计意图、领域知识与自动化硬件生成 ChipGPT-style processor design agents and knowledge-grounded generation flows 构建 ChipGPT 式处理器设计智能体与知识增强生成流程
I study how natural-language specifications, structured microarchitectural knowledge, and automated design flows can reduce the complexity and iteration cost of processor design and verification.
我研究如何利用自然语言规格、结构化微体系结构知识和自动化设计流程,降低处理器设计与验证的复杂度和迭代成本。
02 Dataflow Accelerator Modeling and Compiler Optimization 数据流加速器建模与编译优化 Co-designing models, mappings, compilers, and systems for efficient dataflow execution 围绕高效数据流执行协同设计模型、映射、编译器与系统 LLM-based cost models and compiler-facing design-space exploration 面向编译器的基于大语言模型的代价模型与设计空间探索
My work explores modeling, mapping, compilation, and system-level optimization for dataflow accelerators, with the goal of improving how specialized computation patterns are expressed and executed on hardware platforms.
我的工作探索数据流加速器的建模、映射、编译和系统级优化,目标是改进专用计算模式在硬件平台上的表达与执行方式。
03 Multimodal Generative AI for Hardware Design 面向硬件设计的多模态生成式 AI Evaluating and improving generative models across text, diagrams, and code 面向文本、图示和代码多模态信号评估并改进生成模型 Multimodal Verilog generation and evaluation for synthesizable hardware code 面向可综合硬件代码的多模态 Verilog 生成与评估
I investigate how multimodal signals such as natural language, architecture diagrams, and hardware code can be jointly modeled for hardware design, and develop evaluation methods for generation quality, synthesizability, and reliability.
我研究自然语言、体系结构图示和硬件代码等多模态信号如何联合建模以服务硬件设计,并开发面向生成质量、可综合性和可靠性的评估方法。
- Natural-Language Processor Design Automation 自然语言驱动的处理器设计自动化 ChipGPT-style processor design agents and knowledge-grounded generation flows 构建 ChipGPT 式处理器设计智能体与知识增强生成流程
- Dataflow Accelerator Modeling and Compiler Optimization 数据流加速器建模与编译优化 LLM-based cost models and compiler-facing design-space exploration 面向编译器的基于大语言模型的代价模型与设计空间探索
- Multimodal Generative AI for Hardware Design 面向硬件设计的多模态生成式 AI Multimodal Verilog generation and evaluation for synthesizable hardware code 面向可综合硬件代码的多模态 Verilog 生成与评估
Featured Talk 报告与媒体
Talks and Lectures 讲座报告
National and Official Media Coverage 国家官方媒体报道
Quoted remarks 媒体引述 English excerpt 中文摘录 Read quote 查看引文
“In the view of Kaiyan Chang, a researcher at the State Key Laboratory of Processor Chips, Institute of Computing Technology, Chinese Academy of Sciences, AI is in fact an imprecise form of generation. In many cases, AI may not meet engineers’ design needs, so improving the correctness of AI in chip design remains a bottleneck.”
“在中国科学院计算所处理器芯片全国重点实验室研究人员常开颜看来,AI其实是一个不精确的生成。很多时候,AI不一定符合工程师设计的需要,因此如何提升AI在芯片设计中的正确率是一个瓶颈。”
Quoted remarks 媒体引述 English excerpt 中文摘录 Read quote 查看引文
“Kaiyan Chang, a researcher at the State Key Laboratory of Processors, Institute of Computing Technology, Chinese Academy of Sciences, told China Electronics News that ChatGPT can provide enough reusable code for the chip design industry, offer more innovative code patterns, shorten design time, and improve design efficiency. Because chip design requires a massive amount of code, the industry urgently needs an efficient development tool. Backed by a powerful database, ChatGPT shows strong expressive capability in code generation: with only a task prompt, it can automatically generate corresponding code. In addition, unlike software code patterns, chip design code patterns are relatively fixed, and some code can be generated automatically. Therefore, while traditional automated code generation often provides only fixed code patterns, ChatGPT can provide both fixed code patterns and more innovative ones.”
“In the future, ChatGPT will gradually become a key step in front-end chip design automation workflows, and may evolve into another new paradigm after HLS (high-level synthesis) enabled C-language-based chip design,” Chang said.
“中科院计算所处理器全国重点实验室研究人员常开颜向《中国电子报》记者表示,首先,ChatGPT能够为芯片设计行业提供足够的可复用代码,同时还能提供更多的创新性的代码模式,缩短设计时间,从而提升设计效率。由于芯片行业需要巨大的代码量进行设计,因此急需一个高效的开发工具。背靠强大的数据库,ChatGPT在代码生成方面显示出了优秀的表达能力,只需要给它一个任务提示,就可以自动生成对应的代码。此外,与软件代码模式不同,芯片设计的代码模式相对比较固定,一部分代码可以通过自动化来生成。因此,传统的自动化代码生成模式往往只能提供固定的代码模式,而ChatGPT在提供固定代码模式的同时,还能提供更多的创新性的代码模式。”
“未来,ChatGPT将逐渐演变为芯片前端设计自动化流程中极为关键的一步,在未来很有可能演变成为继HLS(高层次综合)开创C语言设计芯片后的又一种新的范式。”常开颜说。”
Quoted remarks 媒体引述 English excerpt 中文摘录 Read quote 查看引文
“Behind me is ChipGPT, an AI tool that assists engineers in chip design and can greatly improve their work efficiency.”
“我身后的叫做ChipGPT,它是辅助工程师进行芯片设计的一款AI工具,可以大大地提高工程师的工作效率。”
Professional and Technical Media Coverage 领域专业媒体报道
Honors and Awards 荣誉奖励
- Special Prize of President Scholarship, Chinese Academy of Sciences (sole recipient at ICT that year) 中国科学院院长特别奖(当年计算所唯一获奖者) The highest-level scholarship recognition of the Chinese Academy of Sciences for doctoral students with outstanding research achievements. 2026
- Outstanding Graduate of Beijing 北京市优秀毕业生 Municipal-level honor awarded to outstanding graduates in Beijing for excellent academic performance, research achievements, and comprehensive development. 2026
- Special Director Award, Institute of Computing Technology, Chinese Academy of Sciences 中国科学院计算技术研究所所长特别奖 Institute-level special recognition by the Director of ICT, CAS, for outstanding research performance and contribution. 2026
- National Scholarship for Doctoral Students 国家奖学金博士生奖 National-level scholarship recognizing top doctoral students for academic excellence and research potential. 2025
- MICRO 2025 Student Travel Grant MICRO 2025 学生差旅资助 Competitive travel support for students attending MICRO, a top-tier computer architecture conference. 2025
- Lude Investment Doctoral Student Award 虑得投资博士生奖 Doctoral student award supporting promising Ph.D. research and academic development. 2025
- Outstanding Student, National Key Laboratory of Processors 处理器芯片全国重点实验室优秀学生 Laboratory-level recognition for outstanding student performance in processor-chip research. 2024, 2025
- Innovation Award, VeriSilicon Cup Integrated Circuit Design Contest 芯原杯集成电路设计大赛创芯奖 Awarded for AI assistant design innovation and achievement in the VeriSilicon Cup Integrated Circuit Design Contest. 2025
- ICCAD 2024 LLM Verilog Generation Contest Honor Award ICCAD 2024 LLM Verilog 生成竞赛荣誉奖 Honor in an ICCAD contest focused on LLM-based Verilog generation and hardware design automation. Collaborate with CUHK group. 2024
- 61st Design Automation Conference Young Fellow 第 61 届设计自动化大会(DAC)Young Fellow Young researcher recognition associated with DAC, a premier conference in design automation. 2024
- Merit Student Model, University of Chinese Academy of Sciences 中国科学院大学三好学生标兵 University-level model merit student honor, a higher distinction among UCAS student awards. 2024
- Outstanding Student Cadre, University of Chinese Academy of Sciences 中国科学院大学优秀学生干部 University-level recognition for student leadership, service, and organizational contribution. 2023
- Merit Student, University of Chinese Academy of Sciences 中国科学院大学三好学生 University-level merit student award recognizing academic performance and overall development. 2023, 2024
- Outstanding Graduate of Sichuan Province 四川省优秀毕业生 Provincial-level graduation honor awarded to outstanding graduates in Sichuan Province. 2021
- Outstanding Undergraduate Thesis, University of Electronic Science and Technology of China 电子科技大学优秀毕业论文 University-level thesis award recognizing undergraduate research quality and thesis excellence. 2021
- National Scholarship for Undergraduate Students 国家奖学金本科生奖 National-level scholarship recognizing top undergraduate students for academic excellence and comprehensive performance. 2018, 2020
Services 学术服务
Journal Reviewer 期刊审稿
- Reviewer 审稿人 IEEE TODAES IEEE TCAD IEEE Communications Magazine JSA ESWA
Conference Reviewer / Program Committee Member 会议审稿与程序委员会
- Program Committee Member 程序委员会成员 AAAI
- Program Committee Member 程序委员会成员 NeurIPS
Academic Roles 学术与组织服务
- Minister of Academic Affairs 学术部长 Graduate Student Union Institute of Computing Technology Chinese Academy of Sciences
- Organizer 组织者 Shishuo Jiyu Student Forum Institute of Computing Technology
- Experiment Teaching Assistant 实验助教 Integrated Circuit and SoC Design Course UCAS Hangzhou Institute for Advanced Study