Synesthetic Music Visualization
A real-time TouchDesigner system linking music amplitude and roughness to visual deformation, followed by a controlled study of soft and intense conditions.
Qixuan Chu / Computer Science / Toronto
I build clinician-reviewable AI workflows and study how adaptive media responds to people.
I’m a Computer Science Specialist at the University of Toronto, graduating in 2027. Right now I contribute to Remeda and a MedQX literature review. Previously, I spent a year testing enterprise workflows at Veeva and co-authored a Pervasive Health paper.
Pervasive Health 2025 · Springer 2026
Open DOI ↗ 02 12 monthsEnterprise test automation at Veeva
Read experience → 03 12 focused testsCompleted referral-draft backend scope
See the boundary ↓Current work
May 2026 - PresentDraft contribution · backend scope
Remeda — clinician-reviewable referral drafts
I contribute to an applied-AI clinical workflow where generated referral text remains a draft: editable by a doctor, traceable through audit records, and protected by deterministic fallback behavior.
01
Structured generation metadata and PostgreSQL persistence
02
Doctor-only edit, approve, and reject APIs
03
Immutable audit events and duplicate-review protection
04
12 focused tests for the completed backend scope
Scope and limits. The completed draft contribution did not include provider-backed generation, referral submission, or the frontend review interface.
Selected work
Artifacts first. The decisions and limits sit beside them.
The projects below are deliberately not presented as equal: one became a publication, one has a public interactive demo, and one remains a course project with limited public evidence.
Box Office vs. Streaming Services
A three-person D3 data story comparing theatrical performance, streaming adoption, and genre trends across countries and time.
Food Item Classifier
A course project comparing decision trees, KNN, and neural networks on a food-item prediction task.
A note from the work
A failed test and a theory about why it failed are not the same thing.
I learned that distinction while debugging enterprise workflows at Veeva. It now shapes how I approach model comparison, user-study evidence, and generated clinical drafts: preserve the observation, label the hypothesis, and make the next check reproducible.
See the experience timelineContact