Press Release
Korea Deep Learning wins a project to automate 46 types of loan documents at a major financial company with DEEP OCR+
Korea Deep LearningPublished

On May 22, 2025, Korea Deep Learning won the "AI OCR Automation of Unstructured Loan Documents" project from a major Korean financial company. The project applies DEEP OCR+, its VLM-based document AI, to automate the classification and entry of 46 types of application and supporting documents submitted for loan underwriting. Coming about a month after DEEP OCR+ launched on April 10, 2025, the win is the first case of Korea Deep Learning's document agent being supplied to the financial sector.
46 types of documents in a single PDF file
Loan underwriting comes with documents in all sorts of formats, such as agreements, ID cards, certified copies of the property register and proof of income. The financial company received a total of 46 types of application and supporting documents as a single PDF file, and staff separated the documents and entered their contents by hand. The process repeatedly led to missing documents, typos and bottlenecks.
The project aims to replace this manual work with AI. Even documents with complex structures, such as certified copies of the property register, will be automated to the point where AI reads and judges them.
From splitting to system integration: how DEEP OCR+ handles the flow
DEEP OCR+ automatically splits the documents inside a merged PDF and classifies them by type. It then extracts the key data from each document and connects it to the financial company's internal systems. The results are checked on a real-time verification dashboard, and the system runs together with robotic process automation (RPA). Instead of going through and entering documents one by one, staff take on the role of verifying the results on the dashboard.
DEEP OCR+ adds analysis of a document's structure and context to optical character recognition (OCR) to extract key information. With a self-training feature, it is designed to keep improving its processing accuracy and speed as it runs. It supports PDF, HWP and Word formats. The underlying model is a vision-language model (VLM) specialized for document understanding, trained on more than 400 million image and text data items that Korea Deep Learning accumulated over five years.
"For AI to be adopted across industries, it first needs the 'ability to see.'"
— Kim Ji-hyun, CEO, Korea Deep Learning
Korea Deep Learning said that document understanding technology based on visual intelligence can be applied to many automation tasks and will become a key element in speeding up the digital transformation of businesses.
Training-free document AI first shown at AI EXPO KOREA 2025
DEEP OCR+ was demonstrated in depth at AI EXPO KOREA 2025, held at COEX in Seoul from May 14 to 16, 2025. It processes documents in many formats without separate training and can be used right away without data collection or labeling, which keeps the initial burden of adoption low. According to the company, it can be applied to unstructured documents, handwriting and documents that mix multiple languages, and can cut document review time by more than 80%.
At the exhibition, Korea Deep Learning presented use cases by industry, focusing on loan documents in finance, legal contracts and public civil petition documents. This win is the case in which loan document processing in finance turned into an actual project. Korea Deep Learning describes DEEP OCR+ as a next-generation VLM-based solution that grasps even the meaning and context within documents, and says it has lowered initial adoption costs that once ran to hundreds of millions of won to nearly zero.
Widening its reach to insurance, the public sector and manufacturing
Building on this financial sector case, Korea Deep Learning plans to extend its reach to industries with a lot of document-centered work, such as insurance, public institutions and manufacturing.
About DEEP OCR
Korea Deep Learning started out by reading documents. Built on KDL's own document-specialized VLM, DEEP OCR is Korea's first VLM-based OCR solution, reading not just the characters but the structure and meaning of a document together. It handles unstructured documents in every kind of layout without building separate training data, and extracts text and key-value data from visual elements such as seals, signatures, tables, numbers and symbols. From loan documents in finance to administrative documents in the public sector and handwritten waybills in logistics, documents people used to copy out by hand become business data through DEEP OCR. More information is available on the DEEP OCR page at koreadeep.com.
About Korea Deep Learning
Korea Deep Learning (KDL), founded in 2019, is an enterprise AI agent company. Built on KDL Frontier, its own vision-language model (VLM) specialized for documents, KDL makes AI agents that take on and complete the document work of businesses and public institutions. Its main products are DEEP OCR, DEEP Parser, DEEP Agent and DEEP Ops.







