EXPERIENCE

From QA Craft to AI Testing

Experience across test management, automation frameworks, performance, Playwright, and AI-assisted testing

Learning Journey

  1. QA Craft Foundations

    Built solid functional, API, and exploratory testing habits — the judgment that still guides every tool I ship.

  2. Team Leadership Across Stages

    Led QA through ByteDance, SOE, and startup contexts — process, people, and release readiness under different constraints.

  3. Automation & Performance at Scale

    Grew from scripts to frameworks — Playwright, Locust / Go load workers (Locust WebUI), and reusable environments, no longer tied to one language.

  4. AI-Native Tooling

    Turned repetitive QA work into products — multimodal generation, review loops, and exportable deliverables.

  5. OpenClaw Skills & Product Systems

    Packaging ClawHub skills and shipping systems like GlobalPulse and the Docker Hub API Gateway so workflows run beyond demos.

Key Challenges

Database bottlenecks under concurrency

Challenge
High-concurrency account and fund flows hit database saturation — latency spiked before the app layer looked unhealthy.
How I Solved It
Locust / Go load profiles (Locust WebUI), SQL / index review with engineering, and reproducible Docker DB environments for shared diagnosis.
Result
Stabilized critical paths under 500+ QPS pressure and shortened the feedback loop between QA and backend.

Complex account / fund / risk workflows

Challenge
End-to-end paths crossed account, fund, and risk-control systems — UI clicks alone could not prove release readiness.
How I Solved It
Layered Playwright + API automation, shared fixtures, and environment parity so regression covered the real business chain.
Result
Raised automation coverage above 80% on priority flows and cut regression cycle time by about 40%.

Regression efficiency at team scale

Challenge
Manual regression could not keep up with release cadence across ByteDance, SOE, and startup delivery models.
How I Solved It
Framework design, CI/CD quality gates, and reusable suites instead of one-off scripts tied to a single language.
Result
Teams shipped with clearer release readiness signals and 20–35% better response-time outcomes on tuned paths.

Turning AI test drafts into installable products

Challenge
LLM-generated cases looked impressive in demos but failed QA adoption — missing coverage, executability, and a path teams could install.
How I Solved It
Productized multimodal intake, a scored human review loop (Test / Developer / Product Manager), and four run modes: Docker, npm, OpenClaw, and local source.
Result
Shipped the AI Test Case Generator as a flagship installable product with live Docker / npm / ClawHub proof — not a notebook demo.

Tech Stack

AI & Agents

  • OpenClaw / ClawHub skills
  • Multimodal test generation
  • LLM review loops
  • Agent-ready CLI & Docker packaging

Test Management

  • QA leadership across ByteDance, SOE, and startups
  • Release readiness and quality gates
  • Hiring, mentoring, and capacity planning
  • Cross-functional delivery with product & engineering

Automation Testing

  • Playwright / Selenium UI automation
  • API automation (language-agnostic)
  • Locust and Go load frameworks + Locust WebUI
  • Framework design & reusable suites

Containers & Delivery

  • Docker / Compose
  • CI/CD pipelines
  • Domestic & open-source DB images
  • Test environment management

Learning Resources

AI Test Case Generator

Flagship installable AI testing product — from requirements to reviewable cases with a human review loop. Start with the case study or live demo.

Open flagship case study

Trading Assistant Core

OpenClaw / ClawHub skill for technical analysis, signals, position sizing, and risk checks — also available via Docker and pip.

View on GitHub

GlobalPulse

Self-hosted market intelligence and scheduled briefings on Cloudflare Workers — try the live app, then dig into how the system is built.

Open live site

Python QA Practice

Locust performance scripts plus API, Selenium, and Appium examples in one practice repo.

View on GitHub

Go Load Practice

Go load frameworks as workers behind Locust WebUI — scale concurrency in Go, keep the Locust operator UX.

View on GitHub

ClawHub

Discover and install OpenClaw agent skills — including tools I publish for QA and trading workflows.

Browse ClawHub

Current Learning Focus

Agent Skills, Multimodal QA & Product Systems

Right now I am deepening the loop from idea → agent skill / product → packaged system that teams can run:

  • Publishing and iterating OpenClaw / ClawHub skills
  • Multimodal requirement → structured test artifacts
  • Locust / Go load craft behind AI tools
  • Product systems on Workers and APIs

Keep going

Start with the flagship AI Test Case Generator, or keep reading Engineering Notes.

Flagship case studyTry demoEngineering Notes

Contact:

Open for collaboration, consulting, and engineering opportunities.

  • AI Testing tools / ClawHub skill customization
  • QA / SDET consulting and team advisory
  • Test infrastructure / Docker environment enablement
  • Open-source collaboration with InnoNestX
mail@xuxuclassmate.com