[Digital Today reporter Seok Dae-geon (석대건)] Dinotesia showcased its enterprise AI agent platform, Seahorse Cloud, at the AI Expo. Dinotesia, a long-term memory AI and semiconductor integrated solutions company, said on May 8 it took part in AI EXPO KOREA 2026 at COEX in Seoul and conducted a live demonstration of the full Seahorse Cloud process based on real-use scenarios in the financial sector.

Seahorse Cloud is a platform focused on easing recurring bottlenecks that arise when companies apply generative AI to practical work. When a document is uploaded, AI splits the content into semantic units and stores it, then searches for relevant content and answers in line with natural-language queries. Using retrieval-augmented generation, it bases responses on in-house data to reduce hallucinations. Based on search results, it can connect with external systems through the Model Context Protocol standard and expand into an AI agent that handles follow-on tasks.

For environments requiring high accuracy, such as finance, defense and public sectors, it includes a structure combining an ontology-based knowledge graph and retrieval-augmented generation. The ontology reflects relationships among documents, regulations and work data to raise search accuracy and link decision rationales more precisely. This enables companies to manage corporate documents and data as knowledge assets connected to AI rather than simple storage, the company said.

At the exhibition, it was set up so visitors could see at once the automatic vectorisation of unstructured documents such as PDFs, tables and scanned documents, the document search process driven by natural-language queries, and the process in which an agent calls functions through MCP and processes results. In particular, it demonstrated on site how search results and execution logs are recorded, with tasks such as loan screening in mind where decision rationales and audit responses are important.

Seahorse Cloud supports both software-as-a-service and on-premises deployments, and is designed so multiple organisations can use it independently at the same time. It is provided in an Amazon Web Services environment and can be started immediately on the official website without separate infrastructure 구축.

Chief Executive Jeong Mu-gyeong (정무경) said, "The competitiveness of enterprise AI depends on how accurately it understands and uses the meaning and relationships of accumulated data." He added, "In addition to powerful document parsing and meaning-based search, Dinotesia combines a domain ontology-based knowledge graph to provide Seahorse Cloud that enables deeper use of enterprise information assets."