Website: https://erishen.cn | GitHub: https://github.com/erishen | LinkedIn: https://www.linkedin.com/in/erishen/
Summary
Focused on platform engineering and developer tooling, backed by a solid full-stack foundation. At PayPal, built from scratch a standardized Docker test environment system covering 6 language stacks and 13 e-commerce platforms; a three-layer image caching architecture reduced cold-start times from minutes to under 10 seconds; pushed 30+ multi-arch (arm64/amd64) images to JFrog; packaged a self-service startup tool and Makefile unified interface so AI Agents and engineers can spin up environments without knowing any underlying details. In parallel, built an AI-assisted code analysis toolchain with emphasis on root-cause analysis and pattern documentation, growing a shared team knowledge base rather than individual expertise. Tech stack covers Docker, Makefile, Shell, JFrog, Python, FastAPI, React, Next.js, Node.js, TypeScript, Redis, and gRPC.
- Platform thinking — Abstract high-frequency, repetitive, error-prone engineering processes into reusable infrastructure; hide complexity so developers and Agents focus on business logic
- Multi-stack environment management — Independently built a standardized test environment system covering PHP / Java / Ruby / Python / .NET / Node.js across 13 e-commerce platforms
- Multi-arch compatibility depth — Systematically resolved arm64 (local Mac) vs. amd64 (CI / cloud) native extension incompatibilities; three-layer image caching architecture is reusable across projects
- AI toolchain engineering — Built AI-assisted code analysis tools integrating sub-second codebase sampling with multi-model analysis into developer-facing workflows
- Documentation discipline — Every environment issue gets a root-cause analysis and reusable pattern document; outcomes accumulate as team knowledge, not tribal expertise
Work Experience
PayPal
Full-Stack Engineer | Platform & Tooling | 2024.07 – Present
In 2026, primarily built AI payment migration verification infrastructure from scratch, providing standardized multi-stack e-commerce test environments, automated verification pipelines, and AI-assisted toolchains for AI Agents and engineers.
- Designed and implemented a standardized Docker test environment system covering 6 language stacks (PHP, Java, Ruby, Python, .NET, Node.js) across 13 e-commerce platforms
- Built a three-layer image caching architecture (dependency cache → platform compilation → application image) to solve cross-platform native extension multi-arch compatibility
- Packaged a self-service startup tool handling environment variable injection, architecture detection, volume pre-population, and database initialization internally — exposing only a few env vars externally
- Unified platform operation commands (start / stop / reset / test / build) with a Makefile interface, establishing standard engineering conventions
- Built an AI-assisted code analysis tool integrating MCP with multiple models (Claude / GPT-4 / DeepSeek / Gemini); supports rapid sampling analysis of large codebases and automatic Docker configuration generation
- Maintained docker-compose-patterns.md and issues-found.md knowledge bases, cataloguing environment engineering lessons ranked by learning value
- Delivered a 13-platform e-commerce payment test system; pushed 30+ JFrog multi-arch (arm64/amd64) images supporting fully offline cold starts
- New platform onboarding time reduced from a half-day of debugging to 1–2 hours
- Ruby gem cold start reduced from 2 minutes to 10 seconds; Java Maven build reduced from 5 minutes to 10 seconds
- AI Agents can independently complete environment startup and payment flow verification without human intervention on environment issues
- Built 60+ Playwright E2E test cases covering the full payment chain: checkout, authorize, capture, cancel, and refund
- Identified and fixed complex platform issues including OrbStack BuildKit isolated network namespace behavior and VirtioFS nested bind mount silent failures; produced reusable solution documents
- Contributed to Next.js 12→14 and Node.js 18→20 stack upgrades; delivered Customer Identity Platform (CIP) end-to-end
Quantex Information Technology (Shanghai)
Full-Stack Engineer | 2024.01 – 2024.06
Led frontend integration for a Disney resort AI digital-human customer service system; initial integration completed in 1 month.
- Implemented Web-based 3D avatar rendering, motion and expression control using React, Three.js, and react-three-fiber
- Integrated Azure ASR / TTS voice sync; resolved iOS / Android WebGL and audio playback compatibility issues
- Delivered the first working integration of the AI digital-human system within 1 month, enabling the project to proceed to the demo stage
Trip.com Group
Senior Frontend Engineer | 2017.04 – 2024.01
Owned Trip.com international Blog / Destinations content channel development and maintenance; drove team React / Node.js tech migration.
- Built the department's first Node.js SSR content platform, driving the team's migration from .NET to React / Node.js
- Long-term ownership of Trip.com international multilingual content channel; continuously optimized Core Web Vitals and SEO
- Drove team tech stack migration; established SSR / SEO / Hybrid multi-channel delivery
- Led 2–4 person frontend sub-teams; established code review standards, engineering norms, and knowledge-sharing culture
Jiuzhen Network Technology (Shanghai)
Frontend Engineer → Frontend Engineering Manager | 2015.03 – 2017.04
Led React Native cross-platform social app development targeting Android and iOS; managed a 4–7 person frontend team; set up Jenkins multi-branch automated build pipelines.
Actiontec (Shanghai)
Software Engineer | 2009.10 – 2015.03
Led Hybrid mobile app development for a photo-management product; built router embedded Boa WebServer and management interface; collaborated with US-based cross-timezone teams.
ZTE Corporation
Software Engineer | 2008.10 – 2009.10
Built alarm module for a 3G network management platform; implemented alarm task processing and desktop client display.
Gksel Information Technology (Shanghai)
Software Engineer → Technical Manager | 2007.04 – 2008.10
Developed Web projects (rental platform, doctor appointment booking); managed a 3-person technical team.
Jiangsu Jinswei Information Technology
Software Engineer | 2005.04 – 2007.04
Developed and delivered custom enterprise CRM system.
Projects
PayPal — Multi-Stack E-Commerce Test Environment Platform
Designed and built unified test environment infrastructure for the AI payment integration verification team. Goal: any AI Agent or engineer picks up the project and has a running environment within 5 minutes, with consistent behavior across arm64 and amd64.
- Covers PHP×6 + Python×1 + Java×2 + .NET×2 + Node.js×1 + Ruby×1 across 13 e-commerce platforms with 20+ independent Docker environments
- Three-layer image caching: taking Ruby as an example — gem-cache (platform-agnostic gem source) → bundle-cache (per-arch native extension build artifacts) → application image; resolves arm64/amd64 native extension incompatibility
- Offline cold start: npm prebuild-cache (@next/swc, better-sqlite3, etc.) and Maven dependencies pre-bundled into JFrog; containers select by architecture automatically without internet access
- Self-service interface: only a few environment variables exposed externally; all initialization details (image pull, volume fill, credential injection, database init) handled internally by the tool
- Makefile unified interface: standardized start / stop / reset / test / build commands across all platforms; AI Agents can invoke directly
AI-Assisted Code Analysis Tool
Rapid analysis tool for large codebases: samples and analyzes 4,000+ file projects in under 1 second; multi-model collaboration generates code structure insights and Docker configurations.
- Integrated MCP toolchain for codebase structure sampling, tech stack identification, dependency analysis, and config file generation
- Supports Claude / GPT-4 / DeepSeek / Gemini multi-model routing by scenario and cost
- Auto-generates Dockerfile and docker-compose.yml drafts; developers adjust env vars and start
Related personal project: ai-analyze — MCP + Serena-based codebase sampling and analysis tool
PayPal — Large-Scale Merchant Site Scanning
Layered scanning of hundreds of thousands of merchant URLs to identify legacy payment integration sites, supporting business team migration decisions.
- Three-layer architecture: DNS pre-check + HTTP static scan + Playwright deep exploration — progressively filters invalid targets
- Identified thousands of legacy WPS payment integration sites; analyzed integration pattern distribution (auto-submit / encrypted / hosted / cart)
- Integrated three-source threat intelligence (URLhaus / Phishing.Army / HaGeZi); auto-filtered malicious sites
- Produced structured multi-category site datasets directly supporting business team migration prioritization
PayPal — Customer Identity Platform (CIP)
End-to-end ownership: design, development, testing, security audit, and production release.
- Built project structure, page routing, BFF API, Redis caching, and K8s production deployment configuration
- Implemented privacy policy management and user consent tracking meeting PIPL compliance requirements
Quantex — Disney AI Digital-Human Customer Service
- Integrated 3D avatar model: Web rendering, Azure ASR / TTS voice sync, and lip-sync; initial integration completed in 1 month
Trip.com — International Content Platform
Project URL: https://www.trip.com/blog
- Built the department's first Node.js SSR content platform; drove team migration from .NET stack
- Long-term ownership of Trip.com Blog / Destinations multilingual content channels; optimized Core Web Vitals and SEO
Open Source
Ongoing personal experiments in developer tooling, CLI engineering, and platform standardization.
Developer Tools / CLI: lobster, nsgm — OpenAI Function Calling-compatible tool registry + REST API (standardized tool invocation for AI Agents); Next.js + GraphQL + MySQL full-stack code generation scaffold (94+ commits, actively maintained)
Component Registry / Docs: shadcn-registry — self-hosted shadcn/ui component registry + Storybook 9 documentation site; demonstrates platform engineering thinking for internal component distribution and standardization
Education
Target Role
Location: Shanghai preferred; open to remote
Availability: Immediately
Target Industries: Developer platforms, internal tooling, AI toolchains, FinTech infrastructure, DevX engineering