2026-07 — Present
Senior AI & Full-Stack Engineer
Altong · Seoul, South Korea
Multilingual Q&A and content community serving users in five languages.
- Built the AI assistant end to end — SSE token streaming, multi-turn memory and per-turn language matching — isolated behind its own thread pool so the main platform stays unaffected under load.
- Designed atomic Redis quota guards (burst, daily and global spend limits) with signed guest cookies to keep LLM cost predictable.
- Shipped voice input with Whisper speech-to-text and a two-stage content language detector that replaced an unreliable single-pass approach.
- Work spans Next.js and TypeScript on the frontend through Java 17, Spring Boot, MyBatis and sharded MySQL on the backend.



