Skip to content
AETHER
中EN
01Home02Experience03Projects04About05Awards
Employed · open to talkZhu WenhaiShanghai / Jiangxi

AI Full-stack Engineer

AetherZhu Wenhai · Computer Science, 2026

  • RAG
  • Agent
  • Workflow
  • Model routing

Graduated in June 2026 from Nanchang University with a degree in Computer Science. Now an AI full-stack engineer at a smart-parking / smart-community SaaS company, owning three product lines. Fluent in AI coding — using it to write code, and turning the way the team works into tooling.

Read the case filesContact me

Product lines owned

  • P-01Smart Parking
  • P-02Smart Property & Community
  • P-03Digital Asset Custody
Agents & MCP ToolingAgent Runtimes & Multi-Agent OrchestrationAI Gateways & Protocol TranslationSpring Boot + MyBatis-PlusNode.js / TypeScriptContext Engineering & Prompt DesignDesign Fidelity & Responsive TypographyLinux / Docker / NginxDocument Parsing & Multimodal ExtractionEnglish (NECCS)Interaction Motion & Frontend AnimationTypeScript & Frontend ToolingVue 2 / Vue 3Git Workflow & Code ReviewGitea / Monorepo EngineeringGoMobile H5 & VantLinux Server Operations & Security BaselineMathematical ModelingTailscale / Headscale Mesh NetworkingWeChat Mini Programs (uni-app / native)MySQL & Data GovernanceNestJS + Prisma + PostgreSQLPython + FastAPIRAG & Vector RetrievalRedis / MQTT / QueuesTechnical Writing & StandardsTest AutomationAI Coding (Vibe / Spec)Agents & MCP ToolingAgent Runtimes & Multi-Agent OrchestrationAI Gateways & Protocol TranslationSpring Boot + MyBatis-PlusNode.js / TypeScriptContext Engineering & Prompt DesignDesign Fidelity & Responsive TypographyLinux / Docker / NginxDocument Parsing & Multimodal ExtractionEnglish (NECCS)Interaction Motion & Frontend AnimationTypeScript & Frontend ToolingVue 2 / Vue 3Git Workflow & Code ReviewGitea / Monorepo EngineeringGoMobile H5 & VantLinux Server Operations & Security BaselineMathematical ModelingTailscale / Headscale Mesh NetworkingWeChat Mini Programs (uni-app / native)MySQL & Data GovernanceNestJS + Prisma + PostgreSQLPython + FastAPIRAG & Vector RetrievalRedis / MQTT / QueuesTechnical Writing & StandardsTest AutomationAI Coding (Vibe / Spec)
01Now

What I do now

A SaaS company in smart parking, smart communities and parking-asset custody. I own full-stack delivery across three product lines.

2026.08 — Present

AI Full-stack Engineer

Current

Smart-parking / smart-community / digital-asset custody SaaS company

Company name withheld by agreement; available on request

Full-stack owner of three product lines — smart parking, smart property, and digital-asset custody: a Spring Boot multi-module backend, a B-side admin console, and several WeChat mini programs (resident and guard). Requirements are delivered AI-first, from design docs and API design through frontend implementation and automated regression. I also run production: account systems, the security baseline, release rollback, and remote access to the internal Gitea.

Mini programs
2 clients
Resident (uni-app) · guard (native)
Mini-program pages
390+
Five subpackages plus access, charging, invoicing
Admin console pages
190+
Vue 2 + Element UI, permission-driven routing
Backend endpoints
170+
Spring Boot monolith, 1,400+ Java files
Full work history

Product lines

03
  1. P-01

    Smart Parking

    Temporary parking, monthly/short-term passes, space rental and sharing, reservations, barrier-free payment, e-invoicing, and the cloud gatehouse (guard app with live video and barrier control).

    TemporaryMonthlyRentalGatehousePayments
  2. P-02

    Smart Property & Community

    Unit binding and member management, owner/tenant certification, property and parking-space fees, visitor invitations, access control, repairs and inspections.

    Unit bindingCertificationFeesVisitorsAccess
  3. P-03

    Digital Asset Custody

    Custody and managed operation of parking-space assets: custody contracts, lease signing and renewal, deposits and move-out review, revenue-split rules, billing and reconciliation.

    ContractsRenewalDepositsRevenue splitReconciliation

Stack

Spring BootMyBatis-PlusMySQLRedisVue 2uni-appNative WeChat mini programMQTTLakala / WeChat Pay
02AI engineering

Four AI engineering capabilities

Not "uses AI tools" but engineering AI into deliverables — agents, workflows, model routing and applications, each with something shipped.

01

Agent Application Development

hermes kernel · all in the cloud · any device takes over

Code, services and agents all live on a cloud host, and laptop, tablet and phone degrade into web-UI clients: the kernel runs in the cloud and any screen with a browser can take over. One message from your phone dispatches work, checks progress or ships a fix — development stops being tied to one desk.

  • Multi-device control: the laptop uses the Desktop remote-gateway registry, the phone pins the dashboard to its home screen, the tablet becomes a second console — one kernel, one session, no environment to migrate when you switch devices
  • Messages as commands: one line in Feishu or WeChat dispatches work, asks follow-ups or accepts results; the agent runs in the cloud and replies in the same thread — requirements move forward on a commute or a business trip
  • Single-writer control plane: one master owns `state.db` and the task board, other nodes are failover rather than co-writers — co-writing tears sessions and double-dispatches tasks
  • Zero inbound ports: the gateway binds loopback only, reached through a Tailscale tailnet behind auth; Feishu and WeChat ride outbound long connections instead of public callbacks
  • Bypass command layer: a watchdog reads `/api/status` every five minutes and pushes anomalies to Feishu at zero token cost — it still restarts the service and exports a crash cause when every model gateway is down
  • An intermediate state on the way to AGI: models and tools still move fast, so the human keeps judgement, authorisation and acceptance while the agent does execution and bookkeeping — not hands-off automation, but semi-automatic work that can be taken over or rolled back at any moment
02

AI Workflow Orchestration

Standardized pipeline · local tooling

Turning "requirement → design → code → verify → docs" into a fixed track with an artifact at every step, then distilling the platforms I use into local tools — so the agent works the track instead of being re-briefed each time.

  • A standardized pipeline: every step has its artifact and pass criterion — design, fix log and module-state templates; weak evidence goes back, not forward
  • Every claim points at a `file:line`; each change ends with mandatory static checks and explicit activation prerequisites (restart / rebuild / re-login)
  • Local tooling · Modao: the design source as an MCP the agent can call — screenshots, widget trees, assets, tokens, so alignment need not be re-explained in prose
  • Local tooling · Yunxiao: test plans, run status and defects wired through, 179 regression cases execute and write back automatically; the flow is a skill, packaged once and reused
03

Model Gateway & Balance Management

Own gateways · magpie fork · expiry-first routing

From a balance board to three self-built platform gateways, a magpie fork that folds the board back in, and routing driven by batch expiry.

  • Three self-built platform gateways (WorkBuddy / Qoder / Trae) exposing Anthropic Messages, OpenAI Chat Completions and Responses from one process
  • Account pools: pick by remaining quota, tiered cooldowns, failover — and a doctor command that names why each account was skipped
  • A magpie fork, magpie plus: power-meter joins the GUI as a sidecar-driven Balance Board with batch-level expiry
  • routing=expiry orders accounts by the board’s batch expiry, with thin wrappers supplying runsOutAt for Qoder / Trae / CodeArts
04

AI Application Development

RAG · vector retrieval · multi-model backends

Shipping LLM capability to production: document parsing and chunking, embedding and retrieval, cited answers, swappable model backends.

  • E-commerce support RAG Q&A: FastAPI + Vue 3, full pipeline from parse → chunk → embed → retrieve → generate
  • Dual backend — local open models or cloud APIs — switched by config, not by code
  • Answers carry citations, so they can be checked and hallucination risk stays low
  • Docker-packaged delivery: identical environments, one command to start
See what these became4 AI projects · 4 work deliveries
03Method

How the work gets done

"AI full-stack" in practice is four rules: state it, evidence it, ship it, regress it.

  1. 01

    Design before code

    Every feature starts with a design doc: goals, evidence, APIs, data model, blast radius, risks, verification.

  2. 02

    Evidence over assertion

    Claims must point back to a file and line. Unknowns are marked "to confirm" rather than smoothed over.

  3. 03

    Deliverable change

    Every change states what a human must do to see it: rebuild, restart, re-login, recompile the mini program.

  4. 04

    Automated regression

    Scripts for APIs, Playwright for web, the official CLI for mini programs — one command, results written back.

04Case files

Case files

Work delivery, AI engineering and independent projects filed separately — they solve different problems and deserve different standards.

14 files

WorkFeatured01

Production Server Governance & Remote Access to an Internal Gitea

Turning one Alibaba Cloud ECS from "it works" into "someone else can take it over" — scripted accounts and a security baseline, one-command releases with automatic rollback, and internal Gitea reachable from home with zero public ports

Role
Server and account-system administrator (solo)
Scale
1 ECS host · 7 admin scripts · 10 operations manuals
Alibaba Cloud ECSAlibaba Cloud Linux 3SSH / ED25519sudo groupssystemd
View details
All case filesWork 4 · AI 4 · Student 3
05Capabilities

Capabilities

Grouped by category. The figures are my honest read; the red ones are what I use every day.

29 items

AI Engineering

07

Agents & MCP Tooling

96

Package internal capabilities as agent-callable tools, and keep them as reusable skills

Agent Runtimes & Multi-Agent Orchestration

93

A resident kernel, a task board, sub-agent decomposition and human takeover

AI Gateways & Protocol Translation

95

Anthropic / OpenAI / Responses / Gemini collapsed into one entrance, with SSE streaming, account pools and key distribution

Context Engineering & Prompt Design

94

Constraints, evidence and acceptance written into the context — not left to luck

Document Parsing & Multimodal Extraction

89

PDF, web pages, video and design files all become structured input

RAG & Vector Retrieval

92

Parse → chunk → embed → retrieve → generate, with cited answers you can check

AI Coding (Vibe / Spec)

98

Day-to-day delivery is AI-driven — intent-first for exploration, spec-first for shipping

Backend

07

Spring Boot + MyBatis-Plus

96

Endpoints inside a 1,400-file multi-module monolith, with tenant isolation, dual auth and payment callbacks

Node.js / TypeScript

94

Main language of the gateways and toolchains — protocol translation, streaming, CLIs

Go

89

The language behind the gateways and single-binary services — concurrency, SSE, SQLite, Wails desktop

MySQL & Data Governance

91

Multiple data sources, indexes and slow queries, tenant isolation and consistency

NestJS + Prisma + PostgreSQL

86

The other TypeScript backend line — modular layering, migrations and queues

Python + FastAPI

88

AI application backends and tooling

Redis / MQTT / Queues

88

Caching, sessions, rate limits and device command delivery

Frontend

06

Design Fidelity & Responsive Typography

91

From design tokens to pixel comparison — not "close enough"

Interaction Motion & Frontend Animation

87

Scroll, reveal and transition — while respecting reduced-motion

TypeScript & Frontend Tooling

91

Vite + vue-tsc + ESLint as one chain, with types following the contract

Vue 2 / Vue 3

94

Both the legacy admin console (Vue 2 + Element UI) and modern frontends (Vue 3 composition API)

Mobile H5 & Vant

88

Mobile H5 adaptation, pull-to-refresh and in-app browser quirks

WeChat Mini Programs (uni-app / native)

96

Cross-platform and native alike, with a working understanding of subpackages, size limits and review rules

Engineering

06

Linux / Docker / Nginx

89

Ship it yourself — build, containerise, reverse proxy, TLS

Git Workflow & Code Review

89

Branches, commit conventions and change notes you can verify

Gitea / Monorepo Engineering

86

Self-hosted source control, workspaces and pipelines behind one entrance

Linux Server Operations & Security Baseline

90

Account systems, SSH hardening, release rollback and audit sweeps — all scripted

Tailscale / Headscale Mesh Networking

87

Opening internal services to home and the road with zero public ports

Test Automation

91

Three-target regression (API / web / mini program) with results written back to the test platform

General

03

English (NECCS)

87

National English Competition for College Students — National Second Prize

Mathematical Modeling

85

APMCM Asia-Pacific Mathematical Modeling Contest — National Third Prize

Technical Writing & Standards

92

Turning tacit experience into documents someone else can follow — ops manuals, design specs, collaboration protocols

06Dossier

Experience

Work experience

02

  1. AI Full-stack Engineer

    2026.08 – Present

    Smart-parking / smart-community / digital-asset custody SaaS company · AI Full-stack EngineerCurrent

    • Three SaaS product lines share one account, tenancy, permission and payment foundation, serving six roles — driver, guard, lot operator, property manager, asset owner and tenant; one order and one billing rule must stay coherent across four clients
    • Shipped core backend paths inside a Spring Boot 2 multi-module monolith (~170 controllers, 1,400+ Java files): monthly renewal, self-service space-management fees, temporary-parking toggle, payment-record aggregation and callback period extension, with the API contract handed to all four clients
    • Payment overhaul: unified six entry points — temporary exit, monthly renewal, reservations, space-management fees and more — from redirecting to the Lakala cashier into in-app WeChat Pay (main-scan), handling validity extension on callback, channel-accurate status queries, zero-amount orders and distributed locks, with old and new channels coexisting per lot for a gradual cutover
    • Custody loop end to end: rent / deposit / service / property / space-management fees in one order at signing, split by contract on callback; renewal extends the original expiry; move-out passes review before the deposit refund fires; split recipients and ratios configured in the console, and every settled amount traces back to contract, order and payment record
    • Worked across the resident mini program (uni-app, 390 pages, 5 subpackages): home rebuild, monthly-pass lookup and renewal, vehicle certification, Lakala main-scan migration, unified session-expiry handling; moved 1,152 KB of static assets to OSS and recompressed to clear the 2,048 KB package limit
    • Built admin-console screens on the Vue 2 console (190+ pages): gatehouse video and barrier permissions, real-time lane notifications, monthly-pass validity, merged arrears payment, split-recipient and deposit-refund review, with permissions delivered by role
    • Multi-tenant security: phone numbers unique per account domain, cross-tenant and cross-level access limits, four-dimensional authorization (identity × role × app center × tenant), duplicate-plate interception across users
    • Delivery and quality: every requirement starts with a design doc whose claims point at a `file:line`, every change ends with static checks and explicit activation prerequisites; 179 regression cases cover API / web / mini program, write results back to Yunxiao and file defects automatically
    • AI engineering: two toolchains — Modao design-source ingestion (MCP server) and Yunxiao-driven automated regression — wired into daily delivery, so reviews, integration and regression leave executable artifacts rather than chat logs
    Spring BootMyBatis-PlusMySQLRedisVue 2uni-appNative WeChat mini programMQTTLakala / WeChat Pay

  2. Operations Development Engineer (Intern)

    2025.06 – 2026.04

    Shangrao Yajie Construction Engineering Co., Ltd. · Operations Development Engineer

    • Handled daily operations and secondary development of the internal business system (Spring Boot + Vue)
    • Wrote scripts to automate routine ops tasks and boost efficiency
    • Participated in API integration and data report development for business analytics
    • Monitored system performance, diagnosed and resolved lag issues to keep the system stable
    Spring BootVueShell ScriptingMySQL

The internship was my first sustained contact with a live production system; what it taught me became the four rules above.

Education & campus

  1. Nanchang University · Computer Science

    2022.09 – 2026.06

    School of Mathematics & Computer Science · BSc

    • Transferred into CS Class 223 in Feb 2024; graduated June 2026
    • Core courses: Data Structures, Computer Networks, Database Systems, OS, Software Engineering
    Data StructuresNetworksDatabasesOSSoftware Engineering
07Honours

Awards & certificates

All awards
08Contact

Contact

Email is the fastest channel — usually answered within a day. Full-stack work, AI applications, agent tooling, or just a conversation about putting AI into daily engineering.

Emailzhuwenhai2024@163.com
LocationShanghai / Jiangxi · remote OK
StatusEmployed · open to conversations