Aprila presents the latest LLM technology and LLMOps direction at ‘AI EXPO KOREA 2024’ < Company < FOCUS < Text of article

As the use of Large Language Model (LLM) increases in various fields such as general companies and public enterprises, the size of the global LLM market is expected to rapidly expand at a CARG of 35.9% from this year to 2030. do.

Meanwhile, Afreeca (CEO Kang Seon-geun), a company specializing in artificial intelligence (AI) development and cloud management platform, will be holding the ‘International Artificial Intelligence Industry Expo (hereinafter referred to as AI EXPO KOREA 2024)’ at COEX in Samseong-dong, Seoul for three days from May 1. ‘Participate in.

AI EXPO 2024 is a single artificial intelligence event, the largest in Asia and one of the top five artificial intelligence (AI) exhibitions in the world. At this exhibition, Afreeca plans to showcase all end-to-end LLM-related services, from consulting including the private LLM service Gazelle RAG to building an LLM system based on the MLOps platform Cheetah.

Aprila’s private LLM service, Gazelle RAG, applies RAG (Retrieval-Augmented Generation) to provide users with accurate answers without hallucinations, and fine-tunes the embedding model and private LLM to provide answers based on internal knowledge. Its main characteristic is providing accurate answers.

In particular, it can be built in a closed environment to prevent data leakage risks that public institutions and private companies are concerned about, ensuring both data reliability and security. In addition, consulting is provided for the entire process, from learning data construction to fine tuning, to enable development of domain-specific engines for each industry, enabling faster and more effective introduction of LLM.

Cheetah is an MLOps platform that supports the entire process from data management required for AI model development to model development, deployment, and HPC infrastructure operation. It provides both convenience for non-developers to easily develop AI and high-performance functions for experts.

Cheetah, which has already been proven for its performance and stability by being supplied to major public institutions and companies such as the National Police Agency, Gwangju Institute of Science and Technology, and Hyundai Glovis, is expanding to the LLMOps platform by adding LLM-specific functions.

First, starting with the function to distribute and monitor RAG-based LLM models, LLM model fine-tuning functions, prompt management functions, and vector data management will be supported sequentially. Through this, it is expected that companies that want to apply LLM technology to internal security data in a closed environment or companies that do not have professional personnel will be able to easily establish a system to introduce and operate LLM.

Kang Seon-geun, CEO of Aprica, said, “We will actively introduce our technology and vision as an AI specialist company that provides total solutions necessary for the introduction of generative AI, from LLM to LLMOps, to domestic and foreign visitors to this exhibition.”

Meanwhile, Afreeca is developing and supplying not only AI-related services, but also ‘Serengeti’, an all-in-one platform for multi-cloud integrated management (CMP) and cloud-native information system construction and operation (PaaS). Serengeti supports managing various cloud environments, VMs, and containers in a standardized format. Currently, Serengeti is being used by the Digital Platform Government Hub Testbed, Korea Industrial Complex Corporation, Korea Employment Information Service, and Daejeon University.