I'm Doyeon Kim, an Industrial Engineering student building AI-driven automation systems — turning complex workflows into intelligent, self-operating pipelines.
Currently attending SKax Bootcamp (operated by SK), focused on AI engineering, backend architecture, and cloud-native development.
Focus Areas
├── LLM-based Agent Workflows
├── MCP Architecture & Tool Orchestration
├── FastAPI Backend Systems
├── Prompt Engineering for Production
└── Enterprise AX Pipeline Design
🥇 Grand Prize — Koscom AI Agent Challenge 2025
An enterprise-grade risk reporting agent built on LangGraph + Claude MCP.
I designed the Report-Master MCP Server as the core orchestration layer.
| Component | Description |
|---|---|
| LangGraph Workflow | Multi-step reasoning pipeline with conditional branching |
| APScheduler | Cron-based automation for scheduled report generation |
| Human-in-the-loop | Slack approval system before report finalization |
| Risk Engine | Compliance rule-based grading with audit trail |
| Report Generator | DOCX regulatory report output via Claude MCP tools |
Reusable MCP Server infrastructure extracted from K-WON
Three starter templates for building Claude AI + MCP agents at different complexity levels.
| Template | Description | Level |
|---|---|---|
01_mcp-chat |
Single MCP + Claude chat interface | ⭐ Starter |
02_mcp-fullstack |
Dashboard + chat + multi-MCP routing | ⭐⭐ Intermediate |
03_mcp-langgraph |
LangGraph + APScheduler + Slack Human Review | ⭐⭐⭐ Advanced |
K-PaaS Application Competition Submission
An AI counseling service helping adolescents recognize and manage their emotional states.
Combines KoBERT emotion classification with GPT empathy dialogue.
My contributions:
Prompt Engineering
- Designed
밍몽(emotional development bot) — character persona, informal Korean register, 40% empathy / 60% question ratio - Designed self-diagnosis bot — PHQ-9 based 9-question flow, scoring ruleset, 3-tier
self_harm_riskclassification, enforced JSON output
FastAPI Backend
- Emotion-temperature-driven automatic bot switching (development bot ↔ self-diagnosis bot)
- Redis session management + KoBERT emotion analysis trigger
- GPT emotion report generation + weekly pattern analysis
Special Prize — National School Safety Data Analysis Competition
End-to-end public data pipeline: preprocessing → pattern analysis → visualization → policy recommendation.
Focused on data-driven approaches to safety improvement strategy.
Backend
AI / LLM Systems
Data
Currently Exploring
├── Autonomous Reporting Agents
├── Compliance AI Systems
├── Conversational Workflow Orchestration
├── Enterprise AX Architecture
└── Decision-Support Automation Platforms