Sung-Ho ParkAI

Sung-Ho Park is a machine learning infrastructure engineer with 8 years of experience building production ML systems, vector search platforms, and natural language processing pipelines. As Embedding Pipeline Lead at Napkin Deals, he operates the pgvector embedding pipeline that powers AI-driven deal matching and buyer-seller alignment, forming the core ML infrastructure of the platform.

Sung-Ho previously worked at Naver, South Korea's largest internet company, where he was a Staff ML Engineer on the search relevance team. He designed the embedding infrastructure behind Naver's semantic search system serving 40+ million daily active users. His work on hybrid retrieval combining dense and sparse embeddings improved search relevance by 28% and became the foundation of Naver's next-generation search architecture.

Before Naver, Sung-Ho was a Research Engineer at Kakao Brain, the AI research subsidiary of Kakao Corporation. He contributed to open-source multilingual language models and developed embedding fine-tuning methodologies for domain-specific applications in e-commerce and finance. Two of his papers on cross-lingual embedding alignment were published at ACL and EMNLP.

At Napkin Deals, Sung-Ho has built an embedding pipeline that transforms deal listings, buyer mandates, and seller profiles into high-dimensional vector representations. The system enables semantic matching that goes beyond keyword overlap, understanding the intent and context behind every deal. His pipeline processes updates in near real-time and maintains an index of millions of vectors with sub-100ms query latency.