BELLA Enterprise AI Agent and sLLM Suite

Skelter Labs is a Korean conversational and generative AI company founded in 2015 that builds AI solutions which understand enterprise context. It received joint investment from a Kakao Brain and Kakao Ventures consortium in 2017, launched speech recognition in 2019 and machine reading comprehension in 2020, and ranked first on the KorQuAD 1.0 and 2.0 Korean reading benchmarks. It now works with Metanet, which is linked from its site.

The BELLA line covers BELLA QNA, a RAG chatbot that grounds LLM answers in retrieved company documents; BELLA LLM, small language models trained on private data for closed networks; and BELLA Agent, launched in 2024, which plans and calls APIs including the BELLA QNA search engine to complete multi-step tasks. Skelter Labs also offers a no-code customer chatbot builder and real-time Korean STT and TTS for AI contact centers, and delivers use case discovery, legacy integration and maintenance. Cases include KB Kookmin Card, Shinhan Investment Corp., the Korean National Police Agency and a construction company.

• BELLA Agent: AI agent that interprets requests, plans steps and calls APIs to automate complex business tasks.
• BELLA QNA: RAG chatbot that grounds LLM answers in retrieved enterprise documents.
• BELLA LLM: small language models trained on private data for on-premise and closed-network use.
• No-code chatbot builder: customer-facing chatbot creation with operational support.
• Korean STT and TTS: real-time speech recognition and synthesis models for AI contact centers.
• End-to-end delivery: use case discovery, legacy system integration, evaluation and maintenance.

Korean banks, card companies and securities firms in Vietnam can deploy closed-network RAG assistants for staff searching internal regulations and product documents. Construction and engineering firms can automate review of unstructured contract and technical documents. Contact centers serving Korean-speaking customers can use the Korean STT and TTS models.

Regulated organizations can run LLM chatbots and agents on private data inside closed networks, which Skelter Labs has done for card, securities and public sector clients. RAG grounding reduces hallucinated answers. KB Kookmin Card reported a significant rise in event participation and gained chat-log analytics for targeted marketing. One vendor covers models, speech, integration and maintenance.

Banks, securities firms and public agencies want generative AI but must keep data inside closed networks, which drives demand for private small language models and on-premise RAG and agents. Vertical AI that understands industry processes is replacing generic chatbots, and Korean enterprises with operations in Vietnam often extend the same vendors to overseas subsidiaries.

Founded in 2015 in Korea; joint investment from a Kakao Brain and Kakao Ventures consortium in 2017. Launched BELLA QNA in 2023 and BELLA Agent and BELLA LLM in 2024. Ranked first on KorQuAD 1.0 with an F1 score of 95.15. The site links to Metanet, and a 2026 blog describes Skelter Labs and Metanet becoming one team.