This is the document view: the complete resume, everything expanded, nothing to type. The interactive terminal version is at ?view=terminal. Machine-readable copies: /resume.md, /resume.json, and the PDFs linked in the head.
About
AI-native development — Claude Code and similar agents are my main development partner day to day. I write custom tools and skills for them, which is what shortens the loop from a requirement to a deploy.
AI-agent development — I've taken agent systems from zero to one on my own: agent flow design, tool integration, multi-agent orchestration, prompt iteration, run monitoring.
Focus
- AI-native development: Claude Code as my daily driver across requirement breakdown, design, and implementation.
- Custom AI toolchains: MCP servers, agent skills, and automation scripts (issue triage, code review, commit standards) wired into team production workflows.
- Fast-iteration shipping: reliably delivering backend and full-stack products under tight delivery cycles.
Hackathon Awards
- Cypherpunk side track (hosted by Sanctum)
- AI FinTech Innovation & Creativity Competition: Social Group 3rd Place
- XueDao Hackathon: Won 3 tracks from Cathay Financial Holding, Zircuit, BGA
- Encode Solana Winter Build Challenge 2025: 1st place, Best Consumer App on Solana track
AI-Native Development
I treat AI coding agents as first-class teammates: I design the harnesses, tools, and skills they run on, then delegate the mechanical work so I can focus on architecture, API boundaries, and business logic.
- Agent-driven delivery — Claude Code and other coding agents as my primary development partner on a daily basis — driving requirement breakdown, design, and implementation, and compressing the cycle from requirements to deployment.
- Custom MCP servers — Build Model Context Protocol (MCP) servers that give agents first-class, typed access to internal tools, data, and services.
- Agent skills & harnesses — Design reusable agent skills and prompt templates that keep agents consistently aligned with team standards for commit messages, code reviews, and test writing.
- Workflow automation — Automation scripts for issue triage, code review, and commit-message standardization — integrated directly into team production workflows.
Stack: Claude Code, MCP, Agent skills
AI-Agent Development
AI-agent development is the core of how I build — not a side experiment.
- Multi-agent orchestration — 12 PydanticAI agents with Pydantic-schema output contracts and a stateful director agent (Extrovid); Google ADK and the A2A protocol for inter-agent collaboration (LorePack); LangChain agent layers for signal parsing and urgency scoring (DeepFlow).
- Quality, cost & testability — LLM-as-judge review with best-of-N selection, keyframe gates ahead of generation spend, and provider factories behind USE_MOCK_* flags so the whole suite runs offline.
Stack: PydanticAI, LangChain, Google ADK, A2A protocol, FastAPI, TypeScript
Work Experience
Backend Engineer — Star Vaults
2026-03 – 2026-05 · Taipei
Took one complete service module from zero to one (e-commerce platform / internal tooling). Architecture and implementation — worked through the architecture decisions with one or two senior engineers, then led the implementation, API design, data model and deployment. Backend — Node.js / NestJS, Python (FastAPI / Django), PostgreSQL, Supabase, Redis, MongoDB. Frontend — built the matching interfaces in React / Next.js, keeping the data contract consistent across front and back. Process — boilerplate, unit tests and repetitive code went to Claude Code. Wrote a set of prompt templates and agent skills so that commit messages, code review and tests followed team convention.
Tech: Node.js, NestJS, Python, FastAPI, Django, PostgreSQL, Supabase, Redis, MongoDB, React, Next.js
Backend & Blockchain Engineer — Jepun International
2025-09 – 2026-03 · Taipei
Owned several services and modules in a mid-sized product line (3–5 person team, front and back). Backend — designed and built RESTful APIs in Python (FastAPI / Django) and Node.js / Express, along with the database schemas behind them (PostgreSQL, MongoDB, Supabase). Third-party integration — integrated multiple third-party APIs and SaaS services, handling cross-system data sync and error paths. AI / LLM features — built the LLM features in the product, covering agent flow design and prompt tuning. Frontend — built the matching interfaces in React / Next.js, keeping the data contract consistent across front and back. DevOps — automated CI/CD pipelines and containerized deploys.
Tech: Python, FastAPI, Django, Node.js, Express, PostgreSQL, MongoDB, Supabase, React, Next.js, CI/CD
Projects
Extrovid
Solo personal project · AI-native film production · 2026-06 – 2026-07
An AI-native film production system. It runs script, casting, storyboard, per-shot generation, voice and rough cut end to end, and you can talk to a director agent at any step to change something live. My work: multi-agent orchestration across 12 PydanticAI agents with outputs contracted through Pydantic schemas; a director agent carrying 8 stateful tools so the whole pipeline can be driven in plain language; a per-stage state machine so failures resolve to a single stage; LLM-as-judge review with best-of-N selection and a keyframe gate ahead of generation spend; and provider factories with USE_MOCK_* flags so 236 tests run offline.
Tech: Python, FastAPI, PydanticAI, SQLAlchemy 2.0 (async), PostgreSQL, Next.js
Categories: ai, agents, web
GitHub: https://github.com/KCL1104/extrovid
LorePack
Owner / Lead Developer · Personal project · 2026-02
An AI-driven collaborative story world platform. Users build persistent worlds, AI agents generate illustrated chapters, and people collaborate across universes. My work: the overall architecture and the multi-agent collaboration flow; A2A protocol integration and the inter-agent messaging layer; and the frontend interface. Engineering focus: orchestration patterns for multi-agent systems, task delegation between agents, and aggregating their results.
Tech: Google ADK, A2A protocol, Next.js, FastAPI
Categories: ai, agents, web
GitHub: https://github.com/KCL1104/LorePack
DeepFlow
Owner / Lead Developer · Personal project · 2026-01
A personal project built on the idea of a "smart defence system": context-aware scheduling and dynamic priority filter out notification noise and protect deep work time. Frontend — Next.js (App Router) with shadcn/ui: the "Focus Player" and the queue visualisation. Backend — Python FastAPI, Supabase and Redis: priority queue management and multi-platform webhook intake (Slack / Jira / Gmail). Agent — Python and LangChain: signal parsing, urgency scoring, context-aware response generation. Technical highlights: a Redis sorted set (ZSET) gives O(log N) dynamic priority queueing; LangChain orchestrates the LLM calls.
Tech: Next.js, shadcn/ui, FastAPI, LangChain, Redis, Supabase
Categories: ai, agents, web
GitHub: https://github.com/KCL1104/DeepFlow
PinTool
Backend Engineer · Hackathon-winning · Team project · 2025-10 – Present
A Solana-native strategy automation platform. Creators publish on-chain strategies; subscribers pay in USDC on-chain for access. My work: backend architecture, API development and subscription verification logic; Anchor contract integration and on-chain/off-chain state sync; and working with frontend to get the integration contract right.
Tech: NestJS, Supabase, Anchor, TypeScript
Categories: blockchain, solana, web3, backend
GitHub: https://github.com/pintoolx
Demo: https://pintool.fun/
Award: Cypherpunk side track (Sanctum) · AI FinTech Innovation Competition
Education
Bachelor of Science in Computer Science — University of Taipei
2021-09 – 2025-06 · Taipei, Taiwan
Focused on software development, algorithms, data structures, and blockchain technology fundamentals.
Self-Learning Blockchain Development — Independent Study
2021 – Present
Self-taught blockchain developer with a strong foundation in Solidity, Ethereum, and smart contract development.
Blockchain Development Bootcamp — Online Learning Platform
2024-06 – 2024-08
Completed an intensive 8-week bootcamp focused on Ethereum development, Solidity, and dApp creation.
Blockchain Study Group — Community
2025-02 – 2025-06
A blockchain study group hosting monthly meetups to discuss and learn about blockchain technology.