Twigfarm Consortium Receives “Excellent” Rating in 2025 National Hyper-Scale AI Data Initiative

Twigfarm Consortium has been awarded an “Excellent” rating in the final evaluation of the 2025 Hyper-Scale AI Ecosystem Expansion Project (Project No. 8 – K-Stock Content Data Construction), organized by the National Information Society Agency (NIA).

The final evaluation comprehensively assessed project performance across data quality, technical execution, management structure, scalability, and practical applicability. Twigfarm Consortium achieved high marks in all major evaluation criteria, resulting in the top-tier “Excellent” rating.

‍

■ Project Overview – K-Stock Content Data Construction

This national initiative aims to build and publicly release large-scale, high-quality multimodal AI training datasets to support the advancement of hyper-scale AI technologies.

The K-Stock Content Data project focused on constructing culturally contextualized Korean image-text datasets to strengthen AI competitiveness in the era of generative and multimodal AI.

Key deliverables included:

• Over 50,000 curated images
• More than 500,000 Korean captions
• More than 500,000 English captions
• Advanced inference-based metadata incorporating objects, actions, context, intent, and cultural background
• Full compliance with AI-Hub public release standards

The dataset is designed to support multimodal AI model training, image retrieval systems, culturally contextual generative AI applications, and large-scale AI research.

‍

■ Strategic Significance

This project represents more than dataset production; it contributes strategically to the national AI ecosystem.

1. Strengthening Korean Multimodal Data Competitiveness

In the global AI landscape, non-English and culturally contextualized datasets remain limited. The K-Stock dataset addresses this gap by providing high-quality Korean-context multimodal data.

2. Enabling AI Ecosystem Expansion

Through AI-Hub public release, the dataset provides accessible infrastructure for researchers, startups, enterprises, and academic institutions.

3. Data–Model Integrated Design

The dataset was structured with direct consideration for AI model training and performance validation, ensuring practical usability rather than static data accumulation.

‍

■ Twigfarm’s Core Technical Capabilities

The successful execution of this project highlights Twigfarm’s advanced AI data engineering capabilities:

Multimodal Data Architecture Design

• Integrated image-text dataset structuring
• Inference-driven prompt engineering
• Culturally contextual metadata modeling

Automated + Human-in-the-Loop Quality Framework

• Large-scale automated caption generation
• Multi-stage human review and validation
• Iterative 1-Cycle quality inspection process

Model-Based Data Validation

• Image Retrieval model-based performance verification
• Recall-based evaluation metrics
• AI performance-linked data quality validation

Security & Governance Infrastructure

• Structured quality management organization
• Controlled access and logging systems
• Compliance with national AI-Hub standards

‍

■ Key Factors Behind the “Excellent” Evaluation

The “Excellent” rating was attributed to:

• Stable large-scale data construction and delivery
• Robust copyright clearance and de-identification processes
• Advanced inference-based captioning framework
• Iterative and systematic quality verification
• AI model-driven validation methodology
• Clear scalability and ecosystem expansion strategy

The evaluation recognized Twigfarm’s ability to deliver a fully integrated pipeline covering design, construction, validation, and deployment.

‍

■ Future Outlook

Building upon the success of this initiative, Twigfarm will continue to:

• Advance multimodal AI dataset engineering
• Expand culturally contextual AI training data
• Develop industry-driven AI data services
• Support the growth of hyper-scale AI ecosystems

Twigfarm remains committed to strengthening AI data infrastructure and driving innovation in multimodal AI technologies.

‍

■ Contact

For inquiries regarding this project or AI data collaboration:

📩 support@twigfarm.net

‍