Multimodal document intelligence and OCR for legal and enterprise workflows
南京通达海 develops multimodal document understanding systems anchored in OCR and LLM-driven intelligence, with a tech stack spanning CLIP, PyTorch, Java, and SQL databases. Active projects reveal a pivot toward legal document automation and speech recognition optimization, while hiring remains weighted toward engineering (9 roles) and sales (8) — suggesting product-market validation in a narrow vertical rather than platform expansion.
南京通达海 is a 34-person software company based in Nanjing, China, building document intelligence solutions that combine computer vision (OCR models), multimodal AI (CLIP-based), and large language models. The product targets enterprise and legal-services workflows where document understanding, contract review, and compliance automation create friction. The engineering and sales hiring mix indicates a business-to-business go-to-market focused on contracts and document processing.
CLIP, PyTorch, Python, Java with Spring, SQL Server, Oracle, MySQL, Tomcat, plus design tools (Figma, Photoshop, Dreamweaver). Stack reflects computer-vision and backend service architecture.
OCR model development, document understanding LLMs, multimodal legal document intelligence, speech recognition optimization, and NLP foundation services. Recent focus on contract and legal document automation.
Other companies in the same industry, closest in size