Skip to content

About

This page gives the human context behind the site. Start with the proof chain for inspectable work; use this page to understand the operating style, milestones, and career context behind it.

My Journey

A decade of key moments that shaped how I think and work. Newest first.

Reached C2 English proficiency at EF

Jan 2026

After enrolling at EF Education in late 2024, completed the program and achieved C2, the highest proficiency level. English opened access to global engineering communities and a broader professional world.

GrowthEnglish

Breaking free from student thinking

Late 2024

Working full-time exposed how rigid the "student mindset" was. Started actively shedding it — learning to think in business impact, not just technical correctness.

GrowthMindset

Chose industry over academia — full commitment

Sep 2023

Decided to go all-in on industry rather than pursuing postgraduate studies. No plan B. That clarity accelerated everything that followed.

GrowthCareer

Freshman year: adapting to university life

2020–21

First year away from home. Navigated the transition to independent living and built the study habits that powered the next three years.

EducationGrowth

高考: top of class in science

Jun 2020

Graduated as the science top scorer after three years of focused, disciplined study. The result that opened the door to Computer Science.

EducationGrowth

Operating Profile

How I build, what I can prove publicly, and where private work stops.

AI product engineer focused on turning real workflow pain into usable tools. My strongest direction is the intersection of product sense and engineering execution: identify a repeated operational problem, design a lightweight system around it, ship the tool, document the method, and keep improving it with feedback.
  • Built Digital Twin as a file-first operating layer that turns personal knowledge, skills, and writing rules into executable agent workflows.
  • Built CLIProxyAPI to expose OpenAI/Gemini/Claude/Codex-compatible APIs for coding tools, including OAuth login, provider routing, and multi-account load balancing.
  • Built ManageUp as an anti-fluff workplace communication system, packaged as both an npm-published MCP server and reusable skill library.
  • Published Claude Code Sourcemap as a research/tutorial project, translating 1,884 recovered source files into an accessible architecture learning path.
  • Led end-to-end quality design for traffic gateway anti-crawler systems, frontend security frameworks, and circuit breaker / degradation platforms.
  • Defined validation strategies for high-traffic, high-concurrency request scenarios, including abnormal traffic and multi-region data isolation coverage.
  • Built an API + Web UI automation framework that significantly increased core regression coverage and reduced release regression effort.
  • Established requirement audit and release control loops that maintained zero P0/P1 production incidents over an extended period.
  • Championing full-lifecycle AI integration across the SDLC — from feasibility analysis and PRD/TD generation to code review, test design, UAT, launch verification, and post-release maintenance.
  • Built a team knowledge base to reduce LLM hallucinations and improve test case generation accuracy to 90%.
  • Developed an internal AI agent for policy consultation, workflow routing, and data queries, reducing transactional workload by 30%.
  • Designed prompt templates for code review and test generation, improving usable AI output from 50% to 85%.
  • Integrated AI-assisted review and test generation into CI/CD, cutting manual process cost by 30% and increasing incremental code coverage by 20%.
  • Introduced end-to-end release checkpoints from requirement review to merge and production rollout for high-frequency, high-risk delivery environments.
  • Designed a pipeline gatekeeping system combining risk scanning, compliance validation, and automated regression checkpoints with full audit traceability.
  • Scaled the governance model to 5 business lines, reducing rollback rate by 40% and manual audit cost by 40%.
TypeScriptPythonReact / Next.jsNode.jsCI/CDAPI AutomationLLM IntegrationPrompt EngineeringPyTorchQuality GovernanceProduct ThinkingTechnical WritingMCP / Agent Tools
  • Chengdu University of Technology, B.S. Computer Science, 2020 – 2024
  • Ranked 1st in comprehensive major ranking across four years.
  • National First Prize, China University Computer Game Competition; two Lanqiao Cup provincial second prizes.
  • Intermediate Software Test Engineer
  • NCRE-2, Java
  • CEFR C2 English (IELTS 8.0 equivalent)
  • CET-6
  • Problem discovery from repeated workflow friction
  • Writing-first productization: README, docs, tutorials, cases
  • Open-source packaging: npm, MCP, skills, GitHub Pages
  • Cross-team communication and quality system 0-to-1 buildout