Haechan Jung

// AI-Native Builder

I build and ship real products using AI coding agents — Claude Code, bolt.new, Cursor. 4 deployed services, deep learning pipelines with Claude API, a Grand Prize (1st) in AI Competition, and a Sejong Mayor's Award at a hackathon. All within one year of starting to code.

GitHub Email Featured: Allergy Scan
// How I Work

"I start with real problems, not tech stacks. I prototype fast, ship to production, and iterate with real users."

01
Identify a real user problem
02
Rapid prototype with AI agents
03
Ship to production — fast
04
Iterate with real feedback
// Awards

Grand Prize (1st Place)

T-SUM AI Competition | 2025

Voice-based drug identification system using YOLOv8 + ResNet-18 for visually impaired users. Led data pipeline, model development, and system integration.

Sejong Mayor's Award

Chungcheong Hackathon | 2025

Built MVP in 2 days with AI-powered rapid prototyping. Collaborated with planners and designers.

// Skills
Claude Code bolt.new Cursor Claude API Prompt Engineering React TypeScript Next.js FastAPI Python PyTorch YOLOv8 ResNet ONNX Tailwind CSS Vite PWA Docker Vercel HuggingFace Make
// Projects

Allergy Scan

Deployed

AI-powered food allergy safety service for tourists in Korea. Photo recognition + barcode scan + real-time interpreter in 10 languages.

  • Fine-tuned ViT model (150 food classes, 134K images) with ONNX optimization
  • Claude API: allergy risk analysis + barcode fallback reasoning
  • 3-tier fallback: Open Food Facts → Korea Food Safety API → Claude
  • 3 Seoul public datasets (120K+ restaurants, ingredients, foreign population)
  • Total infrastructure cost: $4
ReactTypeScriptFastAPIONNX Claude APISQLiteDockerVercelHF Spaces
Live Demo GitHub

Voice-Based Drug Identifier

Grand Prize (1st)

AI system for visually impaired users — photograph pills, get instant voice identification with drug name and effects via TTS.

  • 2-stage pipeline: YOLOv8n (detection, mAP50 0.993) → ResNet-18 (classification, 95.4% acc)
  • AI Hub public drug dataset: 6,011 train / 5,520 val annotations
  • Fine-tuned with synthetic hand+pill images (Gemini Flash) to close domain gap
  • Drug info DB from public data portal + Web TTS for real-time voice guidance
YOLOv8ResNet-18PyTorchWeb TTSGemini APIPython

SojangSchool

Deployed

Online course platform with payment system. Solo built in 2 weeks with bolt.new.

  • Full course browsing, enrollment, and payment system
  • Shipped in 2 weeks from idea to production
  • Custom domain, live and operational
ReactVitebolt.new
Live Site

RideOrNot

Deployed

Real-time taxi ride-sharing app for university students.

  • Naver Maps API + Kakao Login integration
  • Real-time ride matching (create/join pools)
  • PWA with native mobile UX
Next.jsTailwindNaver MapsKakao Auth
Live App

Real Estate Lead Automation

In Production

Workflow automation for a real estate agent — analyzed manual process, built automated lead pipeline.

Meta Ads → Lead Form → Make → Telegram → Response
  • Identified automation target by analyzing non-developer's workflow
  • Built, delivered, and currently in daily production use
MakeMeta AdsTelegram Bot

Semo

Sejong Mayor's Award

Sejong City hotspot map. MVP in 2 days at Chungcheong Hackathon.

  • 2-day MVP — collaborating with planners and designers
  • AI-powered rapid prototyping led to hackathon victory
HackathonTeamAI Prototyping
// Background

Korea University (Sejong Campus)

B.S. Computer Software | 2024 - Present | Senior

LikeLion

Korea University Sejong | 2025 - Present

AI tool workshops, project development, onboarding non-developers to AI tools

// Let's Talk

Looking for someone who ships fast with AI?
I'm open to FDE, AI Engineering, and remote opportunities worldwide.
Feel free to reach out anytime — I'd love to chat.

gocks77777@naver.com GitHub
built with pretext + vibes — 2026