로고

로고
  • L²(LLM)
  • L²(LLM)
    • L²(LLM)

    임픽스의 발자취를 소개합니다.

    L2(LLM)

    LLM X Layer
    Industry domain-specific language model

    L² is a language model based on Small Language Model specialized for industry-specific domain knowledge.
    It enables knowledge graph-based navigation and automatic generation of regulatory documents in conjunction with internal systems such as MES and ERP, and its lightweight structure makes it easy for even small and medium-sized manufacturing companies to adopt, enabling them to streamline their work and assetize organizational knowledge at the same time.

    Key Functions

    optimizing on-site application by Lightweight, domain-specific language model

    Learn industry-specific terms and knowledge, enabling high-accuracy Q&A even in low-spec environments

    sLM(small Langauge Model)
    Structuring the Knowledge Graph Industry Domain

    Structures industry-specific standard terms and process relationships into a knowledge graph, and enhances domain-based AI inference through visual exploration and automatic expansion.

    Knowledge graph system
    MCP Protocol-based Intelligent Service

    Based on MCP (Model Context Protocol), it connects industry-specific contexts and systems (ERP, MES, etc.) to ensure security and scalability for automatic recognition and conversion of meanings.

    Model Context Protocol (MCP)
    Human-AI Collaboration Agent

    Intelligent AI platform that supports expert-level question and answer, collaboration, and cause analysis through interactive interfaces and explainable AI features

    Human-AI collaboration agent

    Features

    • Microservices
      architecture

      • Independent module composition for scalability
      • Service integration via API gateways
      • Container-based deployment (Docker, Kubernetes)
      • Utilize an inter-service messaging system (Kafka)

    • AI
      frameworks

      • Develop PyTorch/TensorFlow-based models
      • Utilize Hugging Face Transformers
      • Optimize Inference with ONNX Runtime
      • MLflow-based Model Lifecycle Management

    • Data Management
      Technology

      • MongoDB/PostgreSQL Data Storage
      • Elasticsearch-based Search Engine
      • Implementing a Neo4j-based Knowledge Graph
      • Building an Apache Airflow data pipeline

    • MLOps
      framework

      • Automate model performance monitoring and retraining
      • Data drift detection system
      • CI/CD pipeline integration
      • Automate model versioning and deployment

    L2 View

    Groupware unified search

    Unified LLM-based search for documents and messages in groupware, including mail, meeting minutes, work instructions, and more.
    Search based on semantics, not words, so you can quickly navigate to hidden business information.

    Integrate data from timekeeping systems

    Search, summarize, and analyze real-time data by integrating with time system such as ERP, MES, PLM, etc.
    Users can query data and respond to documents in natural language without accessing separate systems.

    Document audit

    Automatically detects and tags key issues, risk factors, and unresolved items in work documents.
    Increase document audit efficiency with tracking audit trail, automatic highlighting, and summary reporting.

    Architectures