South Korean AI startup Upstage is ramping up global expansion in the US and Japan, positioning its full-stack AI platform to serve enterprise clients. With a focus on boosting productivity through proprietary technologies, the AI firm is trying to replicate its domestic success in two of the world’s most advanced AI markets. Co-founded in 2020 by Sung Kim, former AI development lead at Naver, the startup announced its global expansion strategy, aimed at helping countries build “sovereign AI” models based on their business productivity-focused AI technologies, reported Yonhap News Agency. In April last year, Upstage raised 100 billion won ($72 million) from investors including SK Networks and KT Corp to fuel its expansion. Kim said the funding would help accelerate the development of purpose-trained LLMs and strengthen the company’s global presence. Upstage’s strategy is built around its own AI technologies, like the OCR “DP” technology and the large language model “Solar.” Kim pointed out that these innovations are central to their goal of using AI to boost productivity across industries. The company believes its control over the entire AI stack sets it apart. The latest Solar Pro model, with 22 billion parameters, has shown substantial improvements in key benchmarks. Upstage plans to introduce a multimodal AI model in June that will integrate its DP technology and Solar LLM, along with a new inference AI model. Their latest language model, Solar Pro 1.3, claims to have the “highest benchmark performance” among Korean AI, and they’re promising even more muscle with the 33-billion-parameter “Solar Pro 1.5,” expected to arrive in June 2025. In the US, Upstage plans to leverage its expertise in document parsing, especially within industries like insurance, where it has successfully deployed solutions for Korean insurers such as Samsung Life Insurance. Upstage’s DP technology can process documents at approximately 0.6 seconds per page, translating to 100 pages in under a minute. This ability to quickly convert unstructured documents into structured data is especially relevant for US sectors such as legal, healthcare, and finance, which routinely handle large volumes of text. The structured output supports downstream AI applications like summarization, question answering, and intelligent search capabilities increasingly sought by document-heavy enterprises. As part of its global growth strategy, Upstage has also been involved in a project to develop a Thai-language-specific LLM. Syn-ergy in Japan The company, which counts South Korean firms such as Samsung Card and Hanwha Life among its clients, is focusing on Japan’s extensive document-driven enterprise market. To make inroads, the company is working on a Japanese-specific language model with local partners. It has recently appointed Hiroyuki Matsushita, who has held senior roles at Panasonic and AWS, as head of its Japanese operations. A key offering from the company for the Japanese market is “Syn,” a compact language model co-developed with the Japanese chatbot startup Karakuri. Designed specifically for sectors like finance, healthcare, manufacturing, and legal services, Syn aims to enhance operational efficiency for Japanese companies.   Full-stack AI Upstage is pursuing a full-stack approach to AI by developing and managing its core technologies in-house. Its proprietary ecosystem spans the entire AI adoption journey—from data ingestion to model deployment—anchored by its Document Parse (DP) engine and Solar large language model, which are designed to work in tandem to address specific enterprise needs. By building these foundational tools internally, Upstage aims to offer more integrated, customizable solutions, with the added benefit of faster iteration and seamless system integration. To support broader adoption, the company is also organizing developer hackathons in key regions such as the US, alongside its ongoing R&D efforts. In June 2025, Upstage plans to roll out a multimodal AI model that combines its DP engine and Solar LLM to process both structured documents and unstructured text. The launch is part of its broader push to deliver scalable AI solutions tailored for enterprise environments.