Dinotisia Unveils Innovative 'Seahorse' Vector DB and Public AI Solutions at COMEUP 2025

Amidst the grand opening of the global startup festival 'COMEUP 2025', Denotisia, a long-term memory AI and semiconductor solution company, is drawing attention with its innovative technology. Through this event, Denotisia fully unveiled its core vector data infrastructure technology and real-world AI services. The highlight of this exhibition is the next-generation vector database 'Seahorse'. Seahorse is a technology that vectorizes and stores text, images, and videos, and dramatically improves search accuracy by deeply reflecting context. With the recent proliferation of Retrieval-Augmented Generation (RAG) and agent architectures in LLM (Large Language Model)-based services, the importance of data search performance has grown more than ever. Seahorse is strategically responding to this by presenting an optimal solution. To enable companies to easily adopt a semantic search environment without the burden of building separate infrastructure, Seahorse is also offered as 'Seahorse Cloud,' a fully managed SaaS. Denotisia's technological prowess doesn't stop there. It is clearly highlighting its hardware differentiation by presenting a model that significantly reduces latency, power consumption, and cost compared to existing CPU/GPU systems through 'VDPU (Vector Data Processing Unit),' the world's only semiconductor dedicated to vector operation acceleration. At the exhibition hall, the demonstration of 'poliQ,' an AI service based on public data, is receiving an enthusiastic response. poliQ is a service that reconstructs approximately 30,000 records of National Assembly plenary and committee meeting minutes into vector data, clearly presenting relevant statements, context, summaries, and sources in real-time when a user asks a question in natural language. This is an excellent opportunity to directly witness how powerfully Denotisia's semantic search technology operates in a large-scale document environment. The 'DNA' series, Denotisia's proprietary LLM, has been applied to poliQ. They have continuously enhanced model performance by successively unveiling 'DNA-R1,' which outputs Korean inference processes, and 'DNA 2.0,' which strengthens tool calling and agent functions. Denotisia emphasizes that this innovative structure can be successfully expanded not only to the public sector but also to various enterprise environments dealing with vast amounts of information, such as healthcare and finance. Jung Moo-kyung, CEO of Denotisia, stated, "We are actively supporting enterprises' AI transformation (AX) based on differentiated semantic search technology and automated RAGOps," revealing their future vision: "In 2026, we plan to introduce new products capable of efficiently processing all data, going beyond selected data." Denotisia will also participate in the 'Promising Enterprise Venture Investment Briefing' held during the exhibition period to share its mid-to-long-term roadmap in the AI infrastructure sector with investors. Denotisia's vision and technological prowess to build core infrastructure for the AI era are garnering even greater anticipation.