Zhang Yukui / KtaoKtao / Zhang Yukui
Clinical medicine undergrad · I use AI-assisted tools in my spare time
to build things that actually run in the real world
A post-2005 clinical medicine undergrad transformed by AI. During the day I move between the hospital and textbooks; the rest of the time I work alone with AI-assisted coding tools to turn the messy customer service, data, and reports of real businesses into systems that run themselves every day. Rather than writing every line by hand, I care more about figuring out how to make AI-written code actually hold up in the real world.
Scan to chat on WeChat (fastest) · or email 2018430472@qq.com
Featured ProjectsSELECTED PROJECTS
CASE 01
Operations data automation: dashboard + reporting robot
Automated data pulls, metric definition governance, real-time dashboards, scheduled WeCom group distribution, and a hard audit gate. Managers view operational data directly on their phones.
CASE 02
Medical aesthetics WeCom AI customer service
Layered AI + human collaboration, dual vision-model coordinate calibration, process supervision, and health checks. Already in production at real storefronts.
OPEN SOURCE
40,000+ star open-source project: three upstream PRs for a medical module
Added medical publishing policies and a fail-closed disclosure process, a Chinese literature parsing client, and EQUATOR reporting guidelines to Academic Research Skills. After multiple rounds of adversarial review, all three PRs were merged from 2026-08-01 to 08-02, and the umbrella issue was closed by the maintainer as completed.
On stage · EndorsementsON STAGE
Appointed by JD.com (Zhengzhou) Merchant Service Center as a training instructor for the “AI Implementation in Practice” track (2026.4–11), and I also take the stage at local AI sharing events to talk about my real, hands-on practices. Related media coverage →
About me · 2026 journeyABOUT & JOURNEY
I first encountered AI in early 2026—it took two months to go from zero-background Vibe Coding to entering a real enterprise environment. I'm not a classically trained programmer, and right now I can't independently write complex systems without AI. My edge is understanding real business: breaking problems into data, flows, states, and verification criteria, then using AI-assisted coding tools to carry prototypes through development, debugging, deployment, and iteration. When real engineering issues arise—low-level architecture, security, high concurrency—professional engineer review and collaboration are necessary.
Personal Vibe Coding beginnings
Zero-background, AI-assisted coding: from automation scripts to small tools, training myself into someone who can make things independently.
Down the agent rabbit hole
Claude Code, OpenClaw, memory systems, multi-agent collaboration—first make myself a heavy user.
Entering a real enterprise environment
Medical aesthetics customer service, Douyin lead sync, and the Xiaohongshu matrix all kicked off at once—my first direct exposure to real delivery pressure.
From “it runs” to “there’s a method”
Content strategy, decision logs, a list of pitfalls—compressing experience into reusable methodology.
Production awareness kicks in
Production hardening for the customer service system, failure alerts, incident post-mortems—from “player” to “someone who delivers.”
Back to campus for finals
Medicine is my safety net. I went back to school to prepare for exams and paused AI development—keeping my studies on track is itself a form of delivery.
Full-time, back on the road
Full-time over the summer: my personal site launched and was properly engineered, the AI customer service kept iterating in production, and the first batch of engineering retrospectives was written up.
The detailed text version is a self-introduction: Full journey here → Happening now →
Pitfalls & thoughtsLESSONS & THOUGHTS
Lessons and realizations knocked out of real projects get pinned to a wall. Here are a few, the whole wall is here →On mobile, tap a sticky note to expand it.
GuestbookGUESTBOOK
Stop by and pin a note. The homepage shows the latest few; all messages are collected on a separate wall.
Loading…