Bran Lee

Founder/CAIO of Devfiance. I design, build and deploy AI systems and web products end to end.

I started coding at nine and have built games, web products and AI systems from first sketch to deployment. I founded Devfiance, where I build AI automation, chatbots and agents for companies, and I served as AI Principal at LF, designing production AI systems. For landing pages and web services I own planning, design tokens, code, deployment and checks on phone widths. My open source work and tools are public, so claims can be checked.

Ask for a quote

Work

Projects I built or contributed to, each with the role I had.

  1. AFK

    Author

    Skill layer that lets a CLI agent read a web game's DOM state and drive it by keyboard and mouse.

    JavaScript, Chrome DevTools Protocol

    github.com/LilMGenius/AFK
  2. COGGS

    Founder of the company building it; the engine work is a team effort

    AI game engine that writes and tests game rules; the coggs.ai gallery judges its games blind against stock coding agents.

    Node.js, Vercel, LLM agents

    coggs.ai
    COGGS, first screen at coggs.ai
  3. daedalgames.com and coggs.ai landings

    Main committer of the shared landing repository

    Two static landing pages and a waitlist on one shared UI package with night and light themes.

    HTML, CSS, JavaScript, Vercel

    coggs.ai
    daedalgames.com and coggs.ai landings, first screen at daedalgames.com
  4. devfiance.com

    Author and main committer (static site, live-figures API, design gates, 390 to 1920 px checks)

    Company site in Korean and English, rendered from one data registry with prerendering and visual gates.

    HTML, CSS, JavaScript, Vercel, YAML data registry, Playwright

    devfiance.com
    devfiance.com, first screen at devfiance.com
  5. Enterprise AI automation and chatbot delivery

    Lead of delivery at Devfiance

    AI systems integration and AX consulting for companies across several industries.

    LLM, RAG, Kubernetes, CI/CD

  6. gtg-demo-video-editor

    Author

    HTML-based source of a game trailer, rendered to video.

    HTML, HyperFrames

  7. kyro-mate

    Author (hackathon entry)

    Running-buddy matcher that scores pace, distance, time and place against a friend's runs.

    React, Vite, Vercel serverless

    github.com/LilMGenius/kyro-mate
  8. lazycodex and omowright

    Contributor

    Codex memory, planning and verification, and an agent-oriented Playwright.

    TypeScript, Node.js

    github.com/code-yeongyu/lazycodex
  9. oh-my-openagent (OmO)

    Contributor to another maintainer's repository

    Agent harness that plans, delegates and verifies.

    TypeScript

    github.com/code-yeongyu/oh-my-openagent
  10. paperthin

    Author

    Skills that give coding agents reusable working habits.

    JavaScript, Markdown

    github.com/LilMGenius/paperthin
  11. polysona

    Author

    Persona layer that extracts and runs several personas across AI agents.

    HTML, Markdown, LLM agents

    github.com/LilMGenius/polysona
  12. posterpress

    Author

    Generates lecture and seminar posters from one poster.json, output as PNG and a single-file HTML.

    Node.js, HTML, CSS

  13. thon-judge

    Author

    CLI that scores hackathon submissions from JSON, Markdown or text and writes an HTML report.

    PowerShell, Codex plugin

  14. win-hooks

    Author

    Repairs Claude Code and Codex plugin hooks that fail on Windows.

    Node.js, JavaScript

    github.com/LilMGenius/win-hooks

Services and starting prices

A starting price covers the basic scope. The final quote follows a look at your requirements.

ServiceRelated workFrom
Landing pageKRW 208,000on Kmong
Web app developmentKRW 1,840,000on Soomgo
AI automation and chatbotKRW 123,000on Kmong
Data crawlingKRW 35,000on Kmong
Product detail page designKRW 58,000on Kmong
Logo and brand designKRW 54,000on Kmong
Presentation and card news designKRW 31,000on Kmong
Video editingKRW 45,000on Kmong
Korean to English translationKRW 79,000on Soomgo
CopywritingKRW 40,000on Soomgo

Computed from market price samples collected on 25 October 2026.

How a job runs

  1. Talk

    We confirm the goal, scope and timeline through platform messages.

  2. Quote

    You get a quote listing the scope, deliverables, schedule and price.

  3. Payment

    On Kmong work under KRW 1,000,000 is one order. Above that, each phase is its own order with its own deliverable, placed after you check the previous phase.

    On Soomgo a 50% deposit starts the work and the balance is paid at delivery. From KRW 1,000,000 the price is paid in 3 parts.

  4. Delivery

    You receive the result with its source files and usage notes.

  5. Revision

    One round of revisions after delivery is included.

I build with AI tools and check every deliverable myself before handing it over.

Contact

Send what you want built and when you need it. You can also reach me on the marketplaces.

Contact me on Kmong

Korean electronic tax invoices available