Pioneering AI Software Factories | Author, Builder & Enterprise AI Consultant
- Location: Seoul, Korea
- Email: dohyun@roboco.io
- Blog: roboco.io/posts, moreagile.net
- LinkedIn: @serithemage
- GitHub: @serithemage
I am pioneering a practical approach to AI software factories: systems that turn human intent into implemented, verified, and deployed software, then use operational feedback to improve the process. As the founder and principal consultant of Roboco.io, I connect original methods, open-source implementations, and enterprise consulting to help organizations redesign how software gets built.
My work spans Intent Engineering, enterprise adoption frameworks, coding-agent benchmarks, and an executable Software Factory Demo. I am writing The Software Factory, a public draft for software organization leaders, and authored Hands-On Vibe Coding (Hanbit Media, 2025).
Software engineer since 1995, with 30+ years of hands-on development, 13+ years of direct customer engagement since joining Mamezou in 2013, and 14 years working in Japan. Previously an AWS Senior Software Development Engineer and Senior Technical Trainer. I personally lead customer engagements from technical assessment and AI adoption through implementation guidance and knowledge transfer.
I develop the methods, build the machinery, and publish the evidence. My focus is the complete operating system for AI-driven delivery: intent, context, agent execution, verification, authority, and feedback.
- The Software Factory — A Korean-language book in public draft, covering software factories and organizational design for the AI era. Its 20 chapters and 15 patterns address policy, execution, evidence, responsibility, and continuous improvement. Manuscript and research.
- The Software Factory Demo — An open-source implementation connecting GitHub issues, AI implementation, independent AI review, automated checks, risk-based merge gates, and deployment. Explore the live application and factory dashboard.
- Failure-to-rule improvement loop — The demo includes a self-improvement PoC that proposes changes to agent instructions and checks when failures recur. An October 12 rehearsal produced an improvement PR and a verification PR demonstrating detection of the seeded failure. Changes to factory rules require human review; this is a bounded PoC.
- Intent Engineering — A discipline for expressing Why, What, and Not in an evolving intent document, giving agents explicit goals and boundaries.
- Enterprise methods — Vibe Adoption Framework structures enterprise transformation; VDLC organizes delivery from intent through verification and evolution; Vibe Up adapts Shape Up to agent-driven development. Guides are available in English, Korean, and Japanese.
- Coding Agent Benchmark — Public experiments comparing model–harness combinations on autonomous RealWorld backend builds, with test evidence, completion limits, and explicit corrections to earlier results.
I also publish software factory research and implementation lessons, including a four-part study of open-source factory practices: evaluation criteria, merge authority, verification evidence, and turning failures into rules.
Operate a solo consultancy, directly owning customer communication and technical delivery across AI adoption, engineering guidance, and training. Develop and publish software factory methods alongside executable examples and empirical evaluations.
- Software factory research and engineering: Author The Software Factory public draft and build its companion implementation. Explore intent-driven delivery, verification gates, human approval policies, and failure-to-rule feedback loops.
- Enterprise adoption methods: Develop VAF, VDLC, and Vibe Up to connect AI adoption with ownership, governance, delivery practices, and organizational change.
- Environmental engineering: Conduct existing-system analysis and guide technology knowledge transfer, web reimplementation, and AI integration planning for a wastewater treatment design platform. The engagement focuses on helping the customer understand and take ownership of a vendor-developed system.
- Enterprise collaboration SaaS: Provide AI consulting, with the customer relationship expanding into additional technical advisory work.
- Enterprise software and education: Undertake AI transformation consulting and AI education engagements, adapting guidance to each customer's business and technical context.
- Developer enablement: Deliver hands-on enterprise workshops and develop reusable learning materials, workshop platforms, and open-source examples.
- Engineering and automation: Build serverless AI agents, MCP integrations, document conversion utilities, and report linting tools. Develop multi-LLM workflows for Git-based engineering analysis.
- Led development and optimization of AWS Skill Builder using domain-driven design and serverless architecture.
- Served as Acting Manager of a development team in Korea, coordinating team workflows and customer-focused engineering work.
- Delivered technical training in AWS architecture, security, DevOps, machine learning, and serverless solutions; promoted to Senior Technical Trainer in July 2019.
- Received AWS 2017 APAC Training Person of the Year (awarded January 2018).
- Explained complex cloud concepts through technical instruction and practical learning materials.
- Delivered enterprise cloud migration projects for aviation and financial services customers in Japan.
- Introduced Agile practices and automated CI/CD pipelines to improve development and delivery processes.
- Global Cybernet System - Tokyo (2002.06 - 2012.12): Java Developer, Architect & General Manager; enterprise and embedded applications, data/business modeling, process standardization, and project management.
- DotnetSoft (LemonSoft) - Seoul (2000.02 - 2002.03): Web Developer & Game Producer; websites and online games.
- Career beginnings (1995): Professional software development with Perl CGI and HTML during the emergence of the web.
- Software factory engineering: Intent and context design, coding-agent harnesses, independent agent review, risk-based merge policies, verification evidence, and feedback-driven rule improvement.
- Customer enablement: Technical training, enterprise consulting, architecture assessment, stakeholder communication, and knowledge transfer; professional experience in Korea and Japan.
- Languages: Python, TypeScript, Go, Java; Bash scripting and Git-based workflows.
- AI and integrations: MCP server development, multi-LLM orchestration, Claude Code plugins and automation, API-backed developer tools.
- Cloud and operations: AWS Lambda, API Gateway, DynamoDB, ECS/Fargate, CloudWatch, CDK, Step Functions, and SageMaker.
- Development practices: CI/CD, DevOps, domain-driven design, automated code review, reproducible benchmarks, and organizational adoption frameworks.
- The Software Factory Demo: Executable companion to my software factory research, with issue-to-deployment automation and a self-improvement PoC.
- Coding Agent Benchmark: Evidence for choosing coding-agent model and harness combinations under explicit completion and quality constraints.
- AI Doc Ready: TypeScript CLI and library for checking Markdown and HWPX report structure, with reusable writing instructions and machine-readable diagnostics.
- Aurora DSQL Experiments: Twelve documented experiments, including nine scoped MVPs, comparing DSQL with RDS and Aurora PostgreSQL across performance, operational effort, constraints, and cost.
- Serverless OpenClaw: On-demand AI agent infrastructure using Lambda, API Gateway, and DynamoDB, with web and Telegram interfaces; presented at AWS Unicorn Day 2026.
- hwp2md: Go utility converting Korean HWP documents to Markdown for reusable document workflows.
Additional projects and open-source contributions
Solar Code ⭐ 13 stars
AI-powered CLI tool built on Upstage's Solar Pro2 model for intelligent codebase interaction. Features native Korean and English support, advanced code analysis, and extensible MCP integration. Enables developers to query, edit, and analyze codebases through natural conversation.
awesome-vibecoding ⭐ 215 stars
Curated resource collection for Vibe Coding and AI-driven development. Includes best practices, tools, learning paths, and community resources.
AI-DLC ⭐ 10
AI-Driven Development Lifecycle (AI-DLC) methodology Korean translation. Comprehensive framework for integrating AI into the software development lifecycle.
aidlc-cc-plugin ⭐ 3
AI-DLC (AI-Driven Development Life Cycle) Claude Code plugin. Adaptive workflow steering rules that integrate AI-DLC methodology directly into AI coding agents. See also aidlc-workflows for the underlying workflow rules.
roboco-cli ⭐ 13
AI-native development scaffolding system for vibe coding with Claude Code. Generates project structures optimized for AI-driven development workflows with pre-configured CLAUDE.md, agents, and best-practice templates.
vibe-ready-cli ⭐ 3
CLI tool that assesses and prepares codebases for Vibe Coding adoption. Analyzes project structure, configuration, and documentation readiness for AI-assisted development.
Enterprise adoption framework with six perspectives, 26 capability items, and a Prepare → Assess → Pilot → Transform journey. Connects governance, ownership, knowledge, and harness engineering.
Vibe-Driven Development Lifecycle: Intent → Context → Co-Construction → Verification → Ship & Observe → Evolve. Includes playbooks, templates, human decision gates, and a weekly operating cycle.
An adaptation of Basecamp’s Shape Up for agent-driven development, addressing shaping, verification capacity, and a four-day build plus one-day cool-down cycle.
intent-engineering ⭐ 10
Ship intent, not code. A practical paradigm for the age of vibe coding — focused on expressing developer intent clearly so AI agents can execute effectively. Explores how intent-driven workflows reshape the human-AI collaboration model.
Roboco’s Vibe Coding guide proposing practices, conventions, and recommendations for AI-driven development. Provides comprehensive guidelines for effective human-AI collaboration in software development.
vibemap ⭐ 3
VibeMap — a structured learning roadmap guiding non-developers through the mindset, tools, and practices of Vibe Coding. Actively developed to lower the barrier to entry for those new to AI-assisted development.
A curated list of discounted and free programs for student developers. Comprehensive collection of resources to help students get started with software development.
Vibe Coding-based career preparation kit for CS and Management Information Systems majors targeting semiconductor companies (Samsung DS, SK hynix, NVIDIA). Applies AI-driven learning to domain-specific career readiness.
Complete learning ecosystem with 7 repositories including:
- handson-vibecoding-demo - Interactive TypeScript demonstration project
- ToDoList - Learning project for Vibe Coding fundamentals
- TicTacToe - First Vibe Coding project showcasing AI collaboration
- realworld-vibe-coding - Production-grade application example
- vibe-coding-calc-demo - Calculator demo for hands-on learning
- hands-on-vibecoding-slides - Comprehensive teaching materials
Advanced projects demonstrating enterprise development patterns:
- Realworld-serverless-microservice - Migrating real-world apps to serverless microservices
- Realworld-build-from-prd - Building applications from product requirements documents
- DORA AI Guide - Unofficial Korean and Japanese summaries of DORA AI research, with source attribution
- learn-with-ai - A one-hour session on learning with AI for university students
- cau-vibecoding-demo - Starting environment for a university coding-agent demonstration
- vibe-coding-workshop - Workshop materials for introducing Vibe Coding
- vibe-coding-toolkit ⭐ 2 - Toolkit supporting Vibe Coding workflows
- ttimes-vibecoding-conference ⭐ 19 - Live demo Todo app for Vibe Coding workshops
- vibecoding-demo ⭐ 34 - TypeScript demonstration project for Vibe Coding fundamentals
- handson-vibecoding-demo ⭐ 8 - Interactive hands-on Vibe Coding learning project
- realworld-java21-springboot3 - Modern Java backend API implementation
- conduit-flask - Blog platform with Python/Flask and Couchbase
- windsurf-todo-demo - 30-minute live coding demonstration app
- realworld-front - Frontend for Conduit platform
- dsql-experiments - Reproducible Aurora DSQL, RDS, and Aurora PostgreSQL comparisons, plus adoption guidance in English, Korean, and Japanese
- s3-experiments ⭐ 7 - What else can S3 do? Exploring S3 as KV Store, Event Store, RDBMS, and more — with benchmarks and honest tradeoffs
- serverless-realworld-monorepo - Monorepo for serverless implementations
- realworld-backend-serverless-microservice - Monolithic to serverless migration
- AI Doc Ready - Markdown and HWPX report linting, reusable writing rules, and regression checks; published as
aidocreadyon npm - ghx-cli ⭐ 5 - Enhanced CLI for GitHub Projects automation and workflow management
- claude-skills - Claude Code skills collection for code review, API design, test generation, and Korean documentation
- hwp2md ⭐ 119 - Tool for converting HWP (Hancom Word Processor) documents to Markdown
- botmadang-mcp ⭐ 3 - MCP server for Botmadang, enabling AI agents to interact with Korean social platform
- plugins ⭐ 21 - Claude Code plugin collection - Skills, Commands, Agents, Hooks
- vmux - Agent-Friendly Terminal Extension for VS Code, enabling AI agents to interact with terminal sessions
- data-analysis-demo - Stock market analysis with AI workflows
- claude-wiki ⭐ 6 - Agent-friendly codebase wikis generated via Claude Code — no API key required
- Coding Agent Benchmark - Autonomous RealWorld backend experiments comparing model–harness combinations, with reports, quality rules, and correction records
- serverless-autoresearch ⭐ 24 - Parallel evolution pipeline for Karpathy's autoresearch on SageMaker Spot Training (H100). 10x faster with HUGI pattern
- topology-efficient-deep-learning ⭐ 2 - Experiments validating deep learning efficiency using topology/structural mathematics (TDA, tensor decomposition)
- advisor-strategy-experiment - Measurement harness for the Advisor Strategy (strong model advises, weak model executes). Implements and scores a RealWorld backend across 5 arms (N=5) to compare quality uplift and cost tradeoffs
- semiconductor-design - AI agent research project for Report-Grounded Vibe-Coded AutoResearch in open-source DL accelerator design. 3-layer architecture (L1 Process + L2 Substrate + L3 Content) with a graphify knowledge graph
- korea-ai-foundation-model-verification ⭐ 49 - Verification project for Korean government's AI foundation models from 5 institutions (Upstage, NAVER, SKT, NC, LG)
- KSAT-AI-Benchmark - AI benchmarking project for Korean SAT tests
- qunova-protein-analyzer - Protein structure analysis tool
Serverless OpenClaw ⭐ 196 stars
Run OpenClaw AI agent on-demand on AWS serverless infrastructure with Web UI and Telegram interfaces. Cost-optimized architecture targeting ~$1/month, presented at AWS Unicorn Day 2026.
claude-code-demo ⭐ 4
RealWorld (Conduit) implementation - Medium.com clone with React 19 + Express demonstrating Claude Code integration patterns and AI-driven development workflows.
updoc ⭐ 5
CLI tool for Upstage Document Parse API, enabling efficient document parsing and processing through command-line interface.
Curated list of discounted and free programs, tools, and resources available to student developers.
Python solutions for 2017 Korean College Scholastic Ability Test mathematics problems, demonstrating computational problem-solving approaches.
Unofficial study guide for AWS Certified Solutions Architect - Associate exam. One of the most popular Korean-language AWS certification resources with comprehensive coverage of exam topics and practical insights.
AWS Class Resources ⭐ 75
Repository of AWS training materials and resources accumulated from years of experience as AWS Technical Trainer, covering various AWS services and best practices.
AWS AI Study ⭐ 52
Collection of AWS AI/ML learning resources including Jupyter notebooks, tutorials, and hands-on examples for AWS AI services like SageMaker, Comprehend, and Rekognition.
Production-ready CDK implementation achieving sub-10-second autoscaling response for ECS Fargate services. Features high-resolution CloudWatch metrics, RPS-based scaling policies, and comprehensive operational guides. Addresses critical scaling latency issues in containerized workloads.
-
The Software Factory — Software Factories and Organizational Design in the AI Era - Korean-language public draft, 2026; open manuscript
-
Hands-On Vibe Coding - Hanbit Media, 2025
-
14 AWS Architecture Patterns You Can Use Right Away - Hanbit Media, 2020 (Korean translation)
-
Practical AWS Workbook - Hanbit Media, 2017 (Korean translation)
-
I am a Programmer Vol.2 - Hanbit Media, 2017
-
I am a Programmer Vol.1 - Hanbit Media, 2016
- InfoQ QCon Tokyo Committee (2014.12 - 2016.4)
- Producer/Host: I am a Programmer Podcast (2015.3 - 2017.9)
- AWS Korea User Group leader (2016.9 - present)
- Yonsei University, B.S. Mathematics and Computer Science
