Hi, I am Marshall (aka OGtwelve).

Full-stack + AI-agent engineer · Founder of TripKit

5+ yrs full-stack — Spring Boot / TypeScript backend lead, frontend end-to-end on my own. Recently led ggcoin at developerz.ai — a payments platform for the AI / agent economy (MCP agent-token layer, real payments, shipped to K8s). Heavy first-party Claude / Codex / GPT user since 2022, active on LeetCode; also running TripKit, Upwork freelance, and UCR CS mentoring.

About

I am Marshall (aka OGtwelve). For the last 5 years I have been doing full-stack work in Shanghai — Spring Boot / Spring Cloud backends as the lead, plus Next.js / React Native frontends written end-to-end on my own — leading procurement & sales platforms, a collectibles consignment system, realtime GIS simulation, TripKit iOS, and more.

Recently, as core engineer at developerz.ai, I delivered ggcoin end-to-end — a digital-currency wallet & payments platform for the AI / agent economy: a double-entry ledger, an MCP agent-token layer (AI agents mint / hold / spend value), real payments, and production security hardening, shipped to Kubernetes via ArgoCD GitOps. I also built autonomous coding-agent orchestrators (ai-task-master / claude-task-master, published to npm / PyPI).

I have been a heavy AI user since 2022 — always on first-party subscriptions (OpenAI / Anthropic / Codex CLI / Cursor), never on a Chinese proxy provider. From GPT-3.5 to Claude 4, from Cursor to Codex CLI, these are my daily tools. Not just for writing code: prompt engineering, structured output, RAG, function calling, and full Agent patterns — all running in real production. TripKit's AI itinerary generation and editing pipeline is the headline case.

Algorithms aren't a buzzword either: active on LeetCode (CN), worked through UCR CS141 Algorithms + CS170 AI + CS172 NLP on my own. Took national bronze (Henan provincial champion) at the CCCC competitive-programming competition during college.

Right now my energy splits across three lanes: TripKit (solo product), Upwork international freelance, and CS coursework mentoring for UCR students (algorithms, OS, compiler, DB, AI — all on the table).

AI / LLM
  • Claude (Sonnet 4.6 / Opus 4.7)
  • GPT-5 / Codex
  • Cursor
  • DeepSeek
  • Prompt engineering
  • Structured output
  • RAG
  • Function calling
  • Agent / Tool use
  • MCP / agent-token
Algorithms / CS
  • Active LeetCoder
  • BST / DP / Graph
  • A* / CSP / Q-learning
  • word2vec / RNN
  • Decision trees / SVM / NN
Backend
  • Java 17+
  • Spring Boot 3
  • Spring Cloud
  • Node.js / Bun
  • Hono
  • WebSocket
Data / Middleware
  • PostgreSQL
  • MySQL
  • Redis
  • RabbitMQ
Full-stack / Frontend
  • Next.js
  • React Native
  • Solid.js
  • Tailwind CSS
  • TypeScript
  • Canvas
Other languages
  • Python
  • C++
  • Rust
  • C#
  • SQL
Engineering / Ops
  • Docker
  • Kubernetes
  • ArgoCD / GitOps
  • Linux
  • Nginx
  • Git
  • Monitoring / alerting

AI engineering & algorithm chops

Heavy first-party AI use since 2022, never on a proxy. Algorithm fluency backed by LeetCode + national-level competitive programming.

01

AI engineering

Production user since 2022 — tools and patterns kept current.

  • First-party subscriptions: OpenAI, Anthropic Claude, Codex CLI, Cursor (since 2022, no proxy providers)
  • Prompt engineering + structured output (JSON Schema / response_format)
  • RAG: vector store + hybrid retrieval, taken to production
  • Function calling / Tool use with multi-turn planning
  • Agent patterns: Claude Agent SDK, Codex CLI workflow, MCP toolchains + a self-built MCP server
  • Real cases: ggcoin MCP agent-token payment layer (production), TripKit AI itinerary generation, the ai-task-master orchestrator, Codex-assisted large refactors
02

Algorithms — not an interview prop, a daily tool

From a national-level competitive-programming background to active LeetCode practice.

  • Active on LeetCode (CN)
  • CCCC competitive programming: national bronze + Henan provincial champion (2022)
  • UCR CS141 Algorithms: BST / DP / bit manipulation
  • UCR CS170 AI: A* / CSP / Q-learning
  • UCR CS172 NLP: word2vec / RNN sentiment classification
  • UCR EE244 Learning Labs: decision trees / SVM / neural-network comparisons
See my LeetCode profile →

What I am doing right now

Three lanes in parallel — my own product, freelance, mentoring.

Selected work

A mix of my own product, main-job work, and solo deliveries.

Experience

  1. Senior Full-stack Engineer (Remote) developerz.ai May 2026 — Present
    • Core engineer on an AI-agent-economy payments platform — ggcoin.ai, end-to-end from ledger core to production
    • MCP agent-token layer: AI agents mint / hold / spend value, with cascading revocation, replay nonces, and rate-limit hardening
    • Real payments (CCBill + crypto adapters) and chargeback / compliance flows; full Solid.js frontend
    • Autonomous coding-agent orchestrators ai-task-master / claude-task-master (published to npm / PyPI)
    • Shipped to Kubernetes via ArgoCD GitOps, with per-service availability alerting and centralized error monitoring
  2. Senior Backend Engineer Shanghai Yuanhuan Information Tech. Aug 2022 — Jul 2026
    • Owned backend for procurement / sales / inventory / finance modules and the core REST API
    • End-to-end ownership: spec review, design, integration, release, and prod debugging
    • RabbitMQ event bus + Redis caching for async + smoothing — kept stock deduction stable under bursty load
    • Excel / CSV bulk import with field validation, error pinpointing, and transactional rollback
    • Built inventory + cashflow monitoring + alerting for ops visibility and fast response
  3. Backend Engineer China Aerospace Science & Industry Corp. Jul 2021 — Oct 2021
    • Realtime simulation GIS system: ArcGIS + Cesium for 2D/3D terrain and model visualization
    • WebSocket channels driving launch-to-impact animation and live state sync
    • Packaged a reusable map interaction SDK (distance measure, annotation, camera control)
  4. Full-stack Engineer Beijing Kechuang Baigu Information Tech. Apr 2021 — Jul 2021
    • Internal enterprise systems — page and API development, bug fixes, integration docs

Education & self-study

Formal degrees plus a UC Riverside CS curriculum I worked through on my own.

UC Riverside self-study coursework

Core labs and projects I worked through following the UCR CS curriculum. This is both the foundation I rely on for TripKit + Upwork work and the basis on which I tutor UCR students.

Say hi

Whether it is AI engineering, backend freelance, UCR coursework mentoring, or just an indie-product chat — any of these channels works.

Résumé

Full PDF · English & Chinese.