메뉴 닫기
BUSINESS
PRODUCT

AX Solution

E-commerce Solution

AI CX

Digital Transformation

COMPANY
KO

Family Site

  • X2BEE
  • groobee
  • PLATEER IDT
서비스 메뉴

AX Solution

AX Solution

It is an AI platform that supports private and hybrid environments,
enabling enterprises to create new value from generative AI development to operation.

Developed with PLATEER's proprietary AI technology accumulated over many years, it supports the entire AI lifecycle from data management to model development and deployment, collaboration, and monitoring. Built on intuitive workflows and secure governance, it delivers efficiency, performance, and trust all at once.

XGEN

  • Enterprise tailored
    AI for performance
  • Agile response through scalable architecture
  • Easy AI utilization with reduced complexity
  • Establishing a reliable AI operation environment
  • Strategic cost∙performance management
  • Continuous quality enhancement

XGEN Key Features

XGEN standardizes and accelerates innovation in enterprise AI utilization, from development to operation and scaling.

  • Feature 01
    For easy AI service
    development AI Service Generator
  • Feature 02
    Designed for scalability
    and flexibility XGEN LLMOps
  • Feature 03
    Tailored for vertical
    domains XGEN LLM
  • Feature 04
    From development to
    operation One-stop service setup

XGEN Scalability

XGEN swiftly and flexibly integrates not only Private LLMs but also external Public LLMs such as ChatGPT and Gemini.

Private
XGEN Private

Components

  • User
  • XGEN
  • Employee
  • Vector DB / Embeddings
  • Private LLM

Data Flow

  1. The user sends a request (Prompt, Query) to the XGEN system.
  2. XGEN interacts with the Employee, Vector DB, and Private LLM through the internal network.
  3. The Vector DB retrieves relevant information from internal data using embeddings.
  4. The Private LLM generates a response based on the retrieved context.
  5. The final response is delivered back to the user through XGEN.
  • Configuration of internal Private LLM (requiring GPU infrastructure and fine-tuning)
  • Minimizing data leakage to external networks
Hybrid
XGEN Hybrid

Components

  • User
  • XGEN
  • Employee
  • Vector DB / Embeddings
  • Private LLM
  • Public LLM

Data Flow

  1. The user sends a request or query to XGEN.
  2. XGEN interacts with both internal (Employee, Vector DB, Private LLM) and external (Public LLM) systems.
  3. Internal data is retrieved from the Vector DB via embeddings to provide contextual information.
  4. Public LLM contributes general language reasoning or knowledge-based support.
  5. XGEN synthesizes responses from both sources and returns the final answer to the user.
  • Integration/unification with external LLM (e.g., GPT) APIs (cost incurred based on tokens)
  • Utilizing internal data through Retrieval-Augmented Generation (RAG)

XGEN Architecture

The XGEN platform is designed with an architecture that offers scalability and flexibility, allowing it to respond organically to diverse domain environments.

XGEN Platform의 전체 구조를 보여주는 다이어그램

Domain

  • E-Commerce
  • Public institution
  • Service
  • Finance
  • Ect

Use Case AI Apps

Customers

  • AI Search
  • AI Chatbot

Improving commerce operational efficiency

  • Copywriter
  • Customer consultation analysis
  • Campaign product recommendations
  • Automated marketing image and video creation

Internal business system

  • Internal document creation/search
  • AI Code Assistant

XGEN Platform

AI Service Generator

Workflow Canvas

  • AI Chatbot
  • Copywriter
  • AI-API's

Deployment management

  • API integration
  • Embed
  • Webpage

RAG management

  • Embedding model
  • Vector DB
  • Re-ranker

Data Source

Structured data

  • ERP/SRM
  • CRM/MES

Policy/Regulation

Unstructured data

  • PDF, WORD, PPT
  • Email, VOC

Corporate characteristics/culture

LLMOps

  1. Training data management
  2. Pre-training
  3. Fine tuning
  4. Model evaluation
  5. Model saving
  6. Model service

Set Service Settings

  • Configure LLM-based service environment
  • Connect Vector DB
  • Ondemand GPU management
  • Ops monitoring

Foundation Model (LLMs)

Public LLM

  • ChatGPT
  • Gemini
  • EXAONE

Private LLM

  • XGEN LLM
  • Polar
  • Gemma
  • deepseek
  • Qwen

Infra

Cloud

  • AWS
  • Microsoft Azure
  • Google Cloud

On-Premise

  • In-house server
  • Storage
  • Network/Security
Polar

Polar supports companies and services developed by flashlightly protect corporate data, and services based on this, and services based on this.

XGEN LLM

Foundation Model (LLMs)

  • Public LLM
    ChatGPT, Gemini, EXAONE
  • Private LLM
    XGEN LLM, Polar, Gemma, deepseek, Qwen

E-commerce specialized Polar key features

  • Optimizing Korean
    Maximize practical work utilization and response quality through 480 million Korean and e-commerce data-based learning
  • Accurate and Reliable Response
    Ensure reliable responses by reflecting the latest information through RAG-based real-time search and generation
  • Stable Operating Environment
    Combine e-commerce operation expertise with sensitive data filtering and monitoring to ensure both stability and security
  • Efficient Integrated Management
    Integrate UI, embeddings, and vector DB seamlessly to enhance performance and reduce operational costs

Polar(e-commerce specialized sLLM)

Polar(e-commerce specialized sLLM)

sLLM + RAG-Based Question Answering Flow

  1. The user enters a Prompt or Query.
  2. The system sends the Query to the Knowledge Content Repository (RAG).
  3. The RAG searches related documents and returns an augmented context.
  4. The system passes the Prompt + Query + augmented context to the sLLM.
  5. The sLLM generates the final Text Response and returns it to the user.

Components

  • RAG (Retrieval-Augmented Generation)
    - Retrieves relevant information from documents and databases to create contextual input.
  • sLLM (Specialized Large Language Model)
    • Analyzes user queries
    • Validates answers
    • Processes generation requests
    • Generates responses

    → Both pre-trained and fine-tuned models are used together.