Press Release
Korea Deep Learning launches DEEP Agent for Manufacturing to structure manufacturing documents without training
Korea Deep LearningPublished

On September 4, 2025, Korea Deep Learning launched DEEP Agent for Manufacturing, a document solution built for the manufacturing industry. The solution automatically recognizes and structures the unstructured documents found on factory floors, such as work orders, process deviation reports and equipment inspection checklists, whose formats vary widely, without any separate prior training.
The limits of OCR that has to be rebuilt every time a form changes
On the factory floor, every supplier has its own document format, and paper documents are still common. With existing OCR, even a small change in a form meant building a separate template and going through repeated training. Documents that mix tables, images and drawings on a single page, such as drawings, bills of materials (BOM), process manuals and cost statements, were even harder to handle.
DEEP Agent for Manufacturing solves this with document understanding technology based on a vision-language model (VLM). Trained on more than 400 million pages of document data, the VLM grasps not only the characters but the context and structure of a document to extract the key information. It handles handwriting recognition, table interpretation and extraction of information inside images at the same time. This approach began with DEEP Agent, launched in March 2025. DEEP Agent is deployed in under two weeks on average without separate training, and it has cut customers' deployment costs from an initial KRW 300 million to KRW 150 million and build time from 66 days to 10 days.
From 'C/T' in injection molding to 'NG judgment' in inspection, designed to understand process terms
Manufacturing documents use different terms and units in each process. DEEP Agent for Manufacturing comes with a dictionary of terms and units for each process. It reads information with an understanding of shop-floor terms such as 'C/T' and 'injection temperature' in injection molding, or 'NG judgment' and 'deviation' in inspection.
After being checked against the original and verified, the extracted data connects to internal company systems such as enterprise resource planning (ERP) and to reports, and is also used in the work that follows, such as yield calculation and process data verification. The solution aims to turn the 'manufacturing tacit knowledge' built up in documents, such as the experience and working standards of skilled workers, into data. It supports deployment through API, on-premises and SaaS.
A solution born from projects in steel structures, textiles and construction
The solution was developed from Korea Deep Learning's project experience in manufacturing and construction. Its groundwork came from cases including connecting quotations from a steel structure and piping company to ERP, structuring multilingual handwritten documents for a textile company, and extracting information from a construction company's inspection reports.
"DEEP Agent is a document intelligence solution that understands and structures documents and carries out the work that follows."
— Kim Ji-hyun, CEO, Korea Deep Learning
Kim described the launch as the real start of document AI optimized for the work of each industry, and said the company would keep introducing practical AI for each industry. Korea Deep Learning plans to release the 'DEEP Agent for' series in turn for finance, logistics, the public sector and other fields.
About DEEP Agent
Korea Deep Learning makes AI that does not stop at reading documents. DEEP Agent is a document-based AI agent that interprets a document's visual structure and linguistic meaning together to extract the key information, then carries out the whole of the document work, from classification and verification to the hand-off into business systems. Upload a contract and it summarizes the key clauses, flags the risky ones and goes on to register it in the ERP, and it reaches the field in two weeks on average with no separate training. It holds GS Certification Grade 1, Korea's national software quality certification, and is designated for priority purchase by public institutions. More information is available on the DEEP Agent 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.







