Company

Com4In is a company specializing in SI, IT, and solutions, founded by experts in Korea's IT industry. Drawing on business knowledge accumulated at large enterprises and diverse experience from venture professionals, Com4In supports customers with IT infrastructure implementation, system operations, solution delivery, and outsourcing services.

Challenge

Com4In needed an AI OCR framework that could reduce the burden of manual entry and verification for documents such as certificates and contracts, where accurate value validation is critical, while also connecting with existing repositories.

• Verification burden for certificates and contracts

  Com4In's key workflows include certificate upload and contract management. For these documents, verifying whether values in the document match reference data is more important than simple text extraction.

• Need to integrate with existing on-premises repositories

  Because Com4In operates on-premises data centers as well as DB and file environments, the new AI OCR system needed to remain connected to existing repositories. The company required a structure that could link extracted data with existing reference data and make it usable in business systems.

• Need for reference data extraction and comparison logic

  The project scope included reference data extraction and data comparison logic. This shows that the Com4In case went beyond OCR automation, focusing on supporting verification work by comparing AI-extracted data against business standards.

• Requirement for improved extraction accuracy and stabilization

  The interim report identified extraction logic accuracy improvement, technical consultation, stabilization, and future planning as key workstreams. To apply document automation in real business operations, both extraction accuracy and operational stability needed to be secured.

• Need for long-term document data utilization

  The project screen structure included a document knowledge search area based on RAG, with features such as document registration, similar document search, registration history, and knowledge data management.

Solution

DDI designed a document verification AI platform tailored to Com4In's certificate and contract workflows, combining AI OCR, reference data comparison, repository integration, and RAG-based search.

[Data Processing Architecture Based on Key AWS Services]

• Designed an AI OCR processing structure for certificates and contracts

  DDI designed the AI OCR platform around Com4In's existing workflows, including certificate uploads, contract management, and on-premises DB and file environments. After document upload, AI extracts key data and connects the results with business reference data.

• Implemented reference data extraction and data comparison logic

  During the project, DDI implemented reference data extraction and data comparison logic. This enabled AI OCR results to become business-ready data that can be compared and verified against existing reference values, rather than simply being stored as extracted text.

• Strengthened usability through repository integration

  Because repository integration was part of the key project scope, the project connected AI OCR outputs with the existing DB and file-based environment. This was essential to ensure the new AI system could be used within the existing document management workflow rather than as a separate tool.

• Developed a Gen AI extraction module and compared LLM performance

  DDI analyzed the existing solution environment, compared LLM model performance, and developed a generative AI extraction module. AI understands document context and extracts key data required from certificates and contracts.

• Built an AI document processing system based on key AWS services

  • Amazon S3: Stores uploaded documents and files for analysis

  • AWS RDS PostgreSQL 16 Multi-AZ: Manages core operational data, including users, templates, and analysis results

  • Amazon Bedrock Claude Model: Supports generative AI-based document analysis and data extraction

  • AWS Bedrock Guardrails: Strengthens safety and governance controls for AI usage

  • S3 Vectors: Supports vector-based search for document knowledge search based on RAG

  • Amazon SES: Provides integration for notifications and email delivery

• Expanded capabilities with RAG-based document knowledge search

  The system includes RAG-based document knowledge search, with sub-features such as document registration, similar document search, registration history, and knowledge data management. This enables certificate and contract data to go beyond storage and expand into similar document discovery and document-based knowledge utilization.

• Reflected business rules through post-processing scripts

  The template management area includes templates and post-processing scripts. This structure can be used to organize AI-extracted data according to business standards or apply comparison logic and validation rules.

• Included AI chat and operational management features

  Key menus include dashboard, AI chat, document analysis (OCR), template management, document knowledge search (RAG), and administration. The administration area includes user management, usage and cost, feedback management, organization management, audit logs, and environment settings to support stable AI OCR operations in an enterprise environment.

Benefit

With AI OCR, Com4In established a foundation not only for document processing automation, but also for certificate and contract data verification and knowledge utilization.

• More efficient certificate and contract verification

  By enabling AI to extract required data from documents and compare it with reference data, Com4In increased the level of automation in verification work. Reference data extraction and data comparison logic are the key differentiators of this case.

• A foundation for AI OCR connected to existing repositories

  By considering integration with on-premises DB and file environments, DDI created a structure that connects AI OCR results with existing business systems. This helps reduce workflow disruption from new system adoption and extends the existing document management flow.

• Reduced manual entry and review workload

  AI extracts key data from documents, while post-processing and comparison logic support the verification process. This helps reduce simple data entry and repetitive checking tasks, allowing users to focus more on exception review and data quality management.

• Turned document data into knowledge assets

  RAG-based document knowledge search was designed to support document registration, similar document search, registration history, and knowledge data management. This creates a foundation for turning certificate and contract data into searchable, reusable knowledge assets over the long term.

• Cloud-native scalability

  By using services such as ECS Fargate, AWS ALB, Amazon S3, RDS PostgreSQL Multi-AZ, Amazon Bedrock, and S3 Vectors, the platform is structured to respond to future increases in document volume and AI capability expansion.

• Stronger AI operations governance and management

  Bedrock Guardrails, usage and cost management, feedback management, audit logs, and organization management features provide the governance and traceability needed to operate AI systems reliably.