Jiakang Huang
I am Jiakang (Daniel) Huang, a Bachelor of Science student in Computer Science at the University of British Columbia. My research interests lie at the intersection of AI systems, AI agents, LLM compilers, distributed databases, and memory systems.
I am currently working with Henry Chan, a Principal Software Engineer at Huawei Canada's Field Lab, where I contribute to the development of GaussDB's shared-nothing distributed database system and its vector database branch. In parallel, I am conducting independent research on graph fusion strategies in PyTorch Inductor. During my third year, I also served as an undergraduate teaching assistant for CPSC 213, supporting students in computer systems and low-level programming.
My long-term goal is to build sustainable, self-evolving AI agents that can continuously learn, adapt, and improve. My research path reflects this vision: my first paper focuses on natural language processing, aiming to help AI better understand human language; my second paper centers on AI infrastructure, targeting faster and more efficient systems for training and serving models; and my planned third paper will explore AI memory systems, enabling agents to develop stronger long-term memory.
Beyond academics and research, I co-founded iMark, an AI bookmark assistant built on retrieval-augmented generation, memory-based personalization, and end-to-end product design. Outside of work, I enjoy basketball and CS:GO, where I once ranked in the top 5% globally.
Education

University of British Columbia
Bachelor of Science in Computer Science

Peking University
Exchange Student, Computer Engineering

Chengdu Foreign Languages School
High School
Experience

University of Toronto
Research Assistant – Graph Fusion Optimization

iMark
Co-founder & AI Technical Lead [Website]

Huawei Canada
Software Engineering Co-op

UBC Department of Computer Science
Undergraduate Teaching Assistant
Blog
March 31, 2026
Deep Dive into speedup_by_fusion in PyTorch Inductor
A benchmark-driven analysis of PyTorch Inductor's speedup_by_fusion config, its runtime logs, and why register spilling can reject fusions that still help.
March 29, 2026
Deep Dive into fuse_nodes in PyTorch Inductor
A structured walkthrough of how PyTorch Inductor enumerates, scores, and greedily applies graph fusion candidates in fuse_nodes.
