Korea Deep Learning
Category: Computer Vision
A specialized AI company providing 'Deep OCR', a VLM (Vision-Language Model) based solution for digitizing unstructured documents, and 'Polyground', a generative AI platform for 3D assets. Korea Deep Learning was founded in 2019. The company is led by Jihyun (Katie) Kim. Based in Seoul, South Korea. Team size: 11-50. Total funding raised: $9.0M. Latest round: Series A ($9.0M, Jan 2026). Key investors include Translink Investment, SV Investment, Korea Development Bank (KDB), IBK Industrial Bank.
- Founded
- 2019
- Headquarters
- Seoul, South Korea
- Team size
- 11-50
- Total funding
- $9.0M
Value proposition
Overcomes the limitations of traditional OCR by using Vision-Language Models (VLM) to understand the 'context' and 'structure' of unstructured documents (handwriting, complex tables) with 99% accuracy, while also automating 3D asset creation for digital content.
Products and solutions
DEEP OCR (VLM-based document digitization engine), Polyground (AI-powered 3D asset marketplace & generation tool), DEEP IMAGE (Super-resolution image upscaling & restoration), DEEP PARSER (Document structure understanding agent)
Unique value
First in Korea to commercialize a 'Vision-Language Model (VLM)' for OCR, enabling the AI to interpret documents like a human (understanding meaning/context) rather than just matching characters.
Target customer
Public Sector (Government/Logistics), Financial Institutions (Banks/Insurance), E-commerce, and Game Developers (for 3D assets)
Industries served
Logistics & Shipping (Mail sorting automation), FinTech (Document verification), Public Administration (Record digitization), Digital Content/Metaverse (3D modeling)
Technology advantage
Proprietary 'DEEP OCR' engine handles unstructured data (e.g., handwritten invoices, damaged historical records) that traditional template-based OCR cannot process. 'Polyground' reduces 3D modeling time from days to minutes using generative AI.
How they differentiate
Utilizes proprietary Vision-Language Models (VLM) for 'Zero-shot' understanding of unstructured documents (unlike template-based OCR); Offers unique 3D asset generation capabilities (Polyground) alongside document AI.
Main competitors
Upstage, Naver Cloud (Clova), Synapsoft
Key partnerships
Korea Post (Logistics automation/OCR), National Tax Service (Document digitization), Hyundai Capital (Financial document processing), NVIDIA (Inception Program partner), Translink Investment & SV Investment (Series A Investors)
Notable customers
Korea Post (Logistics), National Tax Service (NTS), Hyundai Capital, National Archives of Korea
Major milestones
Secured 12B KRW (~$9M) Series A funding led by Translink Investment (Jan 2026), Achieved 10B KRW cumulative revenue without external equity investment, Selected for TIPS (Tech Incubator Program for Startup)
Growth metrics
Achieved 400% average annual growth (CAGR); Cumulative revenue of 10B KRW (~$7.5M) achieved via bootstrapping prior to Series A.
Market positioning
High-growth challenger in B2B/B2G Document AI; Leader in VLM-based digitization
Geographic focus
South Korea (Domestic), expanding to Global (Asia/Europe)
Patents and IP
Holds over 13 registered AI patents, including technologies for 'unstructured data analysis' and '3D object generation'.
About Jihyun (Katie) Kim
Jihyun (Katie) Kim founded Korea Deep Learning in 2019 at the age of 22 while studying Ceramic Engineering and Software at Kyung Hee University. She was selected as the youngest CEO in the 10th batch of the Youth Startup Academy. She successfully bootstrapped the company, securing over 80 enterprise clients and achieving significant revenue before raising external Series A capital. She has led the development of 'Deep OCR' and 3D transformation technologies.
Official website: https://www.koreadeep.com/