Profile
I work at the intersection of cognitive science, user experience research, and software engineering. I am completing an MSc in User Experience Design at Wilfrid Laurier University (December 2026), hold an earlier MSc in Psychology from Beijing Normal University's State Key Laboratory of Cognitive Neuroscience and Learning, and have spent twelve years building web products, interactive courseware frameworks, and developer tooling.
On the engineering side, I write code by hand and at scale: I am the largest contributor to a 280K+ line open-source CJK typesetting pipeline, where I own architecture and performance, and I direct and audit LLM-generated code as a deliberate engineering practice, not a substitute for one. On the research side, I run mixed-methods studies end to end, from experimental design and survey instruments to statistical modeling in R, and I host a 9-week applied statistics workshop for UX practitioners.
My work is grounded in a simple conviction: understanding how minds work leads to better things being made, whether that is a children's textbook on the science of the brain, a reading tool for kids with ADHD, or an open-source music player used by thousands.
Experience
Architect & Top Contributor · CJK typesetting infrastructure
- Rescued the codebase of tiqian, a CJK typesetting system originated by a designer; now its largest contributor, owning architecture and performance, and auditing LLM-generated code across a 280K+ line pipeline including boring, its underlying Haxe-to-Dart/TS/Rust/Swift transpiler.
- Designed a React Fiber-style frame-scheduling priority queue replacing conflicting ad-hoc optimization layers; all rendering tasks are now coordinated under one scheduler.
- Cut CI build time for downstream sites built on tiqian from 25 minutes to 4 minutes 45 seconds (−81%) via Kotlin + Rust bindings, parallelization, and caching.
Open-source products: Rune & Oasis
Rune (符石聆音): an open-source music player reimagining the Zune aesthetic with modern technology. Built with Rust and Flutter, featuring a 61-dimensional audio analysis system for content-based music recommendation via Chroma Spectrum analysis. A UX design experiment: "What if Zune had continued to evolve?" Available on Steam and Microsoft Store. 990+ GitHub stars, 30+ pull requests, 7+ contributors. Covered by IT之家, XDA Developers, Hacker News, and Softpedia.
Oasis: a voice-first psychological self-discovery platform that treats emotions as dynamic networks rather than static scores. Users record voice journal entries, which are transcribed and analyzed through an expert-developed qualitative coding framework (built collaboratively with 5 PhD researchers in psychology and education), then visualized as emotion network graphs. Grounded in Constructed Emotion Theory, Oasis explicitly avoids clinical diagnostic scales to prevent self-diagnosis, positioning itself as a longitudinal narrative-based tool for sustained self-awareness. Sole developer of a 120K-line codebase, orchestrating LLM-assisted implementation while owning the architecture throughout. The platform is currently in closed beta testing.
Workshop Host
Designed and delivered a 9-week applied statistics workshop for UX practitioners: experimental design, hypothesis testing, linear mixed models, bootstrap, permutation, and clustering, taught with hands-on analysis in R and jamovi.
Neuroscience Expert · Cognitive training product for children's reading abilities (EUP)
Designed cognitive training grounded in cognitive and educational psychology for children with reading difficulties (dyslexia, ADHD). Built Bionic Reading-inspired visual guidance, vocabulary curricula using morphological analysis, and Forgetting Curve-based review systems. Participatory design with children and adults with dyslexia. Core philosophy: the goal is for the child to eventually not need the product.
Teaching Assistant & Curriculum Developer
Delivered an experimental "Science of the Brain" course for elementary students, translating neuroscience and cognitive psychology into interactive, age-appropriate lessons. Subsequently led development of a companion textbook (~99 pages, 12 chapters) spanning brain structure, dopamine and motivation, anxiety, attention, flow, task management, memory, and memorization. Responsible for all textbook compilation, typesetting, activity design, and courseware design, bridging neuroscience research with K-12 pedagogy.
Developer, promoted to Technology Leader within two years
Started as a frontend intern building all interactive interfaces for Paperclip (回形针), a popular Chinese science communication project. Led architecture design, managed cross-functional teams, and built an internal content publishing platform end to end.
Created interactive educational games with PIXI.js and Three.js, machine learning visualizations, WebGL tools, and the core API infrastructure for the Recative System interactive video platform. Applied Bloom's Taxonomy and cognitive principles to course design. To keep WebGL content smooth on low-end mobile devices, rebuilt task scheduling around frame slicing and idle-callback execution, and extended the detect-gpu benchmark dataset through statistical modeling to drive device-adaptive rendering. Later at Nice Stick Technology, re-engineered 5 legacy products for performance and internationalization.
Graduate Work · Wilfrid Laurier University
My research at Laurier centers on human-centered LLM products: how people trust, manage, and think with AI systems, and how those systems should be redesigned around human cognition.
Human-Centered AI Research
Trust in ChatGPT across Five Languages
Mixed methods · 181 analytic respondents · Service Safari · Team project (UX607)
A five-language study of how users trust, verify, and restrain themselves around ChatGPT. Found that 43.6% of respondents perceived the system as opaque and 47.0% reported actively restraining their own behavior when using it, while accuracy was the top selection criterion for 66.9%. Produced current-state and target-state service blueprints that locate trust breakdowns at specific touchpoints and propose concrete redesigns for transparency and user control.
Conversation Management in LLM Chat Apps
Mixed methods · Survey and 7 semi-structured interviews · 3 languages
Investigated how users manage accumulating conversation history across LLM platforms. Found that expert users report higher retrieval frustration than novices, not from lack of skill, but from managing 50+ conversations on interfaces designed for casual use. Key finding: user agency (willingness to organize) was a far stronger predictor of satisfaction (p=.05) than technical proficiency (p=.84), while even strict "one topic per chat" discipline failed to reduce retrieval abandonment (p=.85), proving that manual strategies cannot overcome structural UI flaws. Proposed design interventions including response-level bookmarks, frequency-based sorting, and contextual segmentation prompts. Developed the "Management Agency Index," a composite metric isolating active management behavior from technical skill.
Usability & Agent Experience (AX) in jamovi
Mixed methods · Think-aloud protocol · PURE Walkthrough · Agent Experience (AX)
A three-phase usability and AI-interaction study of the open-source statistical platform jamovi. In Phase 1 (user testing, n=6), developed a 6-dimensional qualitative coding framework executed across 3 LLMs with 15 independent passes (agreement 0.84–0.89) to quantify think-aloud protocols; uncovered a critical paradox between low perceived cognitive load and substandard usability (NASA-TLX: 19.8 vs. SUS: 62.9, with 70 error events). In Phase 2, mapped error clusters to a 14-step PURE cognitive walkthrough with 3 UX experts, discovering that invisible cognitive failures in data-type recognition cascaded into severe downstream operation errors. In Phase 3, benchmarked Agent Experience (AX) across 7 frontier systems via Computer Use, demonstrating that interface design directly governs AI reliability: embedding a 2-line textual constraint boosted statistical reasoning accuracy from 56% to 84% across 10 LLMs.
Computational Thinking as Meta-cognitive Strategy for LLM-Assisted Academic Writing
Conceptual framework paper · Human-AI pedagogical collaboration
Synthesizes Cognitive Load Theory (CLT), Problem Space Theory, and Computational Thinking (CT) to address "executive cognitive offloading" in generative AI writing. Argues that unguided LLM use bypasses the desirable difficulties essential for schema construction, fostering meta-cognitive laziness ("performance without learning"). Proposes a systematic pedagogical scaffold mapping 4 CT dimensions across the writing process: decomposition and pattern recognition in ideation, abstraction and algorithm design in drafting (prompting as algorithm design, inverted teaching via the discipleship effect), and logical debugging in revision. Reframes LLMs from cognitive substitutes into cognitive scaffolds, cultivating strategic cognitive offloading so learners retain cognitive ownership.
Design & Build Projects
A Floating 90 · XR Narrative Experience
Godot · Blender · TypeScript · Goxel · Course project
An XR narrative environment reconstructing a 1990s family residence from original blueprints as an enterable archive for intergenerational memory. Designed 38 interactive voxel artifacts and engineered a custom animated shadow baking pipeline: precomputed sparse Cycles ray-traced shadow deltas, packaged them via an automated glTF extension tool, and dynamically blended them in Godot runtime shaders to achieve high-fidelity contact shadows on standalone XR.
L.E.D. · Local Engagement Display, Brantford Smart City
Pitch, Mid/High-Fi Prototyping, Usability Test · Project Site
Community interactive display for downtown Brantford. Full design cycle from concept to usability testing with culturally diverse participants. Identified critical issues in emergency assistance design, navigation accessibility, and physical ergonomics.
Rune Player · UX Case Study
Survey (n=31) and 7 semi-structured interviews
A comprehensive UX evaluation of the Rune open-source music player, combining quantitative survey data with in-depth semi-structured interviews to assess usability, aesthetic perception, and the emotional experience of music browsing. The study applied thematic analysis to interview transcripts and identified key tensions between nostalgic design language and modern usability expectations, producing actionable design recommendations that directly informed subsequent development iterations.
Publications
Affiliative bonding between teachers and students through interpersonal synchronization in brain activity
Social Cognitive and Affective Neuroscience, Vol. 15 (1), 2020
Lifen Zheng, Wenda Liu, Yuhang Long, Yu Zhai, Hui Zhao, Xialu Bai, Siyuan Zhou, Kanyu Li, Huan Zhang, Li Liu, Taomei Guo, Guosheng Ding, Chunming Lu
Contribution: developed the multi-computer experiment software coordinating fNIRS hyperscanning data acquisition; the framework was reused across the lab's later studies.
Shared neural representations of syntax during online dyadic communication
NeuroImage, Vol. 198, 2019
Wenda Liu, Holly P. Branigan, Lifen Zheng, Yuhang Long, Xialu Bai, Kanyu Li, Hui Zhao, Siyuan Zhou, Martin J. Pickering, Chunming Lu
Contribution: built the "dynamic slider" experiment program integrating several external systems, and participated in material preparation and experiment execution.
Neural mechanism of the pair-bonding process in romantic love
Poster, IDG/McGovern Institute & Brain Center Joint Conference, Tsinghua University, 2019
Kanyu Li, Yuhang Long, Siyuan Zhou, Lifen Zheng, Hui Zhao, Yu Zhai, Faxin Zhou, Jinglu Chen, Xinran Xu, Miao Zhong, Ying Li, Chunming Lu
Contribution: data modeling and analysis in R (wavelet coherence, computational cognitive neuroscience modeling) and the research report.
The Contemporary Student Survival Handbook
Independently published in Taiwan, December 2023 · Featured on the sspai Podcast
Kanyu Li (sole author) · A systematic treatment of learning psychology, emotional regulation, and educational critique. Published in Traditional Chinese.
Writing
Certified Author & Top Contributor on sspai.com, one of China's leading technology and digital-lifestyle publications. Selected Matrix-featured articles:
Beyond Grades: Methods, Values, and Meaning Related to Learning
When Will the Homunculus in the Bottle Awaken: People, Artifice, Intelligence, and AI
If Zune Were Still Alive, What Would It Look Like?
The Next Step in Education
AI Won't Eat You
To the Untraceable You Behind Generative Content: On the Ethics of AI-Assisted Creation
Speed Reading Is Worse Than Not Reading: Against the "AI Study Guide" Trend
Reshaping My Relationship with Devices: An Unconventional Phone Experience
When It Sheds the Rhythm Game Skin: A Fit Boxing 3 Review
Also contributed to Jandan (jandan.net) and iPlaySoft (iplaysoft.com).
Selected presentations spanning neuroscience methods (fNIRS/fMRI data analysis), technical workshops (R, presentation design, applied statistics), LLM integration best practices, and public lectures on education and technology. Delivered in Mandarin and English; select recordings available.
Awards & Certifications
Senior Secondary School Psychology Education Teacher Certification
National teacher qualification (PRC), certifying competence to deliver psychological health education at the senior high school level.
UX Upskill Workshop Series
Completed a 4-week workshop covering Canva, rapid prototyping, and 3D printing.
2026Exceptional Contribution Author (2026 Q2) · sspai.com
2025Outstanding Contribution Author (2025 Q2) · sspai.com
Entrepreneurship competition awards relate to Project Bota, a VR-based interactive sandplay therapy solution (LI Kanyu served as Team Leader throughout):
2016National Silver Award · "Challenge Cup" / AVIC National Entrepreneurship Competition (Project Bota)
2016Provincial Gold · Jilin Province Entrepreneurship Competition (Project Bota)
2015Provincial Gold · Jilin Province 1st "Internet+" Innovation & Entrepreneurship Competition (Project Bota)
2015Provincial Second Prize · Jilin Province Extracurricular Academic S&T Works Competition (research on values and subjective wellbeing, analyzed in R with Tidyverse)