At Crossnode LLC, I designed and developed Korean-language generative AI assistant for local information.
I built its LLM question-answering pipeline across query rewriting, routing, parallel retrieval, context construction, generation, and groundedness evaluation. I process Korean public documents and structured data, integrate LLM APIs and retrieval systems, and test responses for factual accuracy, relevance, consistency, and user-intent alignment.
I use Python, FastAPI, Next.js, TypeScript, BGE-M3, reranking models, and LLM APIs to improve reliable AI-assisted question answering. I also review generated code manually and bring hands-on experience auditing Korean-language datasets and evaluating linguistic naturalness, factuality, and instruction following.
