About Me

I'm an AI Engineer focused on building production LLM systems for legal and business document intelligence. I studied Computer Software at Hanyang University, graduating Summa Cum Laude with early graduation, and have been working in applied AI since 2023.

My path into AI was sparked by winning CLOVA AI RUSH and discovering CLOVA Note's AI summarization, which predated ChatGPT. That drew me into NLP, and an internship at NAVER Cloud building CLOVA Note's title generation model set my direction toward applied AI. My graduation project on memory-efficient LLM fine-tuning (combining INT8 quantization with LoRA to train a 13B model on a single consumer GPU) added foundational model optimization experience.

At BHSN, I work across the full stack of legal AI, from model training and evaluation to serving infrastructure and product integration. My work spans contract review AI, LLM serving optimization (vLLM, structured output, long-context handling), evaluation systems combining traditional metrics with LLM-as-a-Judge, and workflow/agent platforms that put AI capabilities directly into legal professionals' hands.

I contributed to predibase/lorax, a multi-adapter LLM serving framework that enables multiple fine-tuned models to run on limited GPU capacity. I fixed chat completion serialization, added deterministic generation via seed parameter, enabled stream usage accounting, and extended structured output compatibility. These contributions directly fed back into our internal serving infrastructure for legal document processing.