AI engineering / platform engineering / senior full-stack engineer focused on landing AI in real engineering workflows.

Previously built customer-facing FinTech systems at PayPal with React, Next.js, and Node.js. Recent focus areas include AI-assisted engineering workflows, system security hardening, and large-scale framework upgrades.

I care less about one-off AI demos and more about systems that are runnable, reproducible, testable, and explainable.

🔍 Open to opportunities

• Landing AI in real software systems (AI-assisted engineering workflows)
• Agent workflow design: planning / execution / independent evaluation / evidence-driven iteration
• Quality & security engineering: reproducible workflows, deterministic checks, and security hardening
• Developer tooling: CLI workflows, prompt / memory systems, and MCP-based code context
• Full-stack systems with React, Next.js, Node.js, TypeScript, Python, FastAPI, Redis, and APIs

📍 Shanghai, China

🌐 Blog: https://erishen.cn

A full AI Agent runtime and engineering workbench on the Cordis DI container — LLM backend / tools / Agent loop / approvals / skills / frontend are all plugins; switching models or adding tools is config-only. Built-in PSE (Planner/Specialist/Evaluator) three-role verification loop and harness orchestration, integrating Makefile scripts, MCP tools and OS-level sandbox (macOS Seatbelt / Linux bwrap). Part of the resolve-* toolchain: resolve-harness (Python/LangGraph skeleton with Fast-path Plugin Runtime — zero model calls for deterministic problems), resolve-tui (Rust CLI/TUI coding Agent), resolve-skills (Agent Skills open standard, consumable by Claude Code / Codex).

An async RAG document pipeline in Rust / axum: upload, chunking, embedding, vector write and hybrid retrieval (BM25 + vector) in one service; a seven-layer architecture decouples storage (SQLite exact dedup + perceptual hash) through retrieval, LLM re-ranking to generation. Local-first and privacy-oriented, built on classic algorithms + heuristics rather than heavy ML dependencies.

An AI BI workbench: connect SQLite / CSV / PostgreSQL / MySQL, ask in natural language, LLM writes read-only SQL and renders ECharts dashboards. A deterministic Text2SQL pipeline (schema constraints + read-only execution + result validation) runs in parallel with a ReAct dual-path (tool calls + reflection/retry), balancing control and exploration; pre-execution permission / scope checks for self-service BI.

Evolved from a single RBAC auth module into a small microservice system: Eureka discovery + Config + Gateway (PEP + hand-rolled zero-dependency JWT) + auth / rbac services, extended with CRM, audit and a BFF layer into a reusable enterprise backend skeleton.

A debuggable C++ Electron digital human workbench — porting the firefly 3D virtual human (React + three.js / R3F) to Electron, with the main process loading C++ native modules via node-addon-api: audio energy / VAD / microphone capture / ASR / TTS wrappers. Integrates Whisper.cpp local offline speech recognition, Piper local speech synthesis, RNNoise noise suppression, and music playback; supports setting breakpoints and debugging C++ directly in VSCode (Node debug main process + lldb attach).

I write about AI engineering, agent frameworks, and full-stack practice on my blog: erishen.cn.

• Evolution of a Distributed Order System: Spring Boot + React from Monolith to High Concurrency — How a Spring Boot + React order system evolves from monolith to distributed, handling high concurrency.
• datapulse: Deterministic Text2SQL Pipeline + ReAct Dual-Path — Architecture breakdown of the AI BI workbench.
• rag-task-service: A Seven-Layer Design for a Rust RAG Service — An async RAG document pipeline from SQLite to LLM re-ranking.
• spring-rbac: From RBAC Auth to a Small Microservice System — Spring Cloud + zero-dependency JWT + CRM + audit + BFF.
• cicdkit Engineering Practice: A Local CI/CD Single-Binary Platform in Pure Go — Building a local CI/CD platform with Go standard library only, single binary deployment.

AI Engineering
Agent workflows, tool calling, MCP, RAG, evaluation loops, prompt systems, memory systems, SSE execution logs.

Platform Engineering
Docker, Makefile, multi-stack environments, reproducible workflows, local tooling, developer productivity.

Full-Stack Systems
React, Next.js, TypeScript, Node.js, Python, FastAPI, Redis, GraphQL, REST APIs, Playwright.

FinTech
Customer-facing systems, identity verification, compliance workflows, secure full-stack development.

• PayPal — Senior Full-Stack Engineer (2024 – 2026)
  Customer-facing FinTech systems, identity verification, AI-assisted engineering workflows, full-stack development with React/Next.js/Node.js
• Trip.com Group — Senior Frontend Engineer (2017 – 2024)
  International content platforms, Node.js SSR, SEO, multilingual content, frontend engineering evolution
• Earlier — Software Engineer → Technical Manager (2005 – 2017)
  Enterprise systems, mobile apps, embedded web, cross-team delivery

• Website: https://erishen.cn
• GitHub: https://github.com/erishen
• Email: erishen@qq.com