Wonsub Song (Tony)
AI-First Full-Stack Developer
Summary
I focus on understanding real customer problems quickly and turning them into systems that are stable and maintainable. My background spans Sports Science and programming, which trained me to ramp fast in unfamiliar domains and structure the real constraint before building.
I designed and templated a modular-monolith architecture (DAEMON-ONE) and an enterprise-style RAG pipeline (DAEMON-KP), then used them to build and operate productized AI systems in a solo development environment.
Communication & Thinking
Asian Parliamentary Debate — University-level champion
University-level Asian Parliamentary Debate champion. Repeatedly trained to analyze unfamiliar topics and structure persuasive arguments within 15-minute preparation windows across economics, policy, healthcare, and technology. WUDC participant. Founded SEDA and ESC and led both.
English: Professional working proficiency.
Experience
Backend & AI Engineer — Intern
Built LawNGood, a legal-dispute discovery and monitoring proof of concept for a client review workflow. Designed and implemented a structured pipeline across collection, extraction, classification, grouping, and review.
Solo · Product Engineer / Architecture Builder
Designed and templated DAEMON-ONE and DAEMON-KP. Built reusable AI system foundations for solo development, deployment, and operation, then applied them across personal projects.
Architecture
DAEMON-ONE — Django architecture template
After building 15+ side projects, I identified repeating pain points in maintainability and onboarding cost. I designed DAEMON-ONE as a structured template embodying deliberate architectural tradeoffs:
- Modular Monolith over microservices — single deployable unit, domain boundaries enforced via
interface.py. - HTMX + SSR over React — server-driven state, zero JS build step, LCP-first rendering.
- Vertical Slicing — logic, template, and style colocated per feature. Delete a folder = feature gone cleanly.
DAEMON-KP extends the same framework into a reusable RAG layer covering ingestion, hybrid retrieval, and citation-grounded answers.
Selected Projects
LawNGood
Legal news monitoring POC for LawAndGood, combining Naver News API collection, article parsing, and Gemini-based classification.
DAEMON-KP
Knowledge-platform RAG architecture covering ingestion, hybrid retrieval, grounded answers, and citations.