This project aims to transform the process of scientific discovery through the use of generative AI to accelerate data analysis, advance inference, and make researchers more efficient in their search for knowledge.

Toward Innovation in Data Analysis, Inference, and Knowledge Creation with AI
Japan has long been a world leader in fields such as physics, materials science, and medicine. This initiative is intended to carry on and develop that tradition as a science and technology nation into the AI age, and to keep Japan at the forefront of data-driven intellectual discovery.
The project also plans to utilize "Syn Pro," a large-scale Japanese language model jointly developed by Upstage and Karakuri, as its base model. Syn Pro" is a high-performance Japanese LLM that is ranked first in Japan in terms of Japanese language performance and is also ranked in the world's top 20 on the Nejumi Leaderboard operated by Weights & Biases.
Toward the New Frontier of Science and AI
In recent years, data in scientific research has been rapidly increasing and diversifying, making it increasingly difficult to extract useful findings from the vast amount of information. In particular, many current systems are unable to "see" unstructured information such as graphs and charts, making it difficult to evaluate data quality. This limitation also hinders the efficiency of analyzing, verifying, and sharing research results, leading to lost industrial innovation opportunities in Japan. In this collaboration, the three parties will combine their strengths to take on the challenge of resolving this issue.
Upstage utilizes large-scale language models (LLMs) and Document Parse technology to achieve advanced information extraction and feature analysis of research data.
Professor Okada has demonstrated theoretical knowledge and leadership in research.
Karakuri Corporation promotes solutions through AI implementation and engineering technology.
The three parties will work together to establish a new AI research model that combines automation, transparency, and collaboration, and will consider ways to secure intellectual property rights for the results.
Professor Masato Okada, The University of Tokyo Comments
We strongly believe that the widespread implementation of LLM (Syn Pro), which has been jointly developed by Karakuri and Upstage in Japan, in addition to Upstage's Document AI, which has an AI-Ready data infrastructure, can solve the essential problems that research institutes and private companies have been facing for a long time. We strongly believe that the widespread implementation of LLM (Syn Pro), which was jointly developed by Karakuri and Upstage in Japan, in addition to AI-Ready data infrastructure and Upstage's Document AI, has the potential to solve essential problems that research institutions and private companies have been facing for many years.
Furthermore, my mission is to build an environment in which this tripartite AI effort can demonstrate its true value in a wide range of industrial domains, including healthcare and manufacturing.
With AI for Science still in its infancy, we look forward to working together to pioneer a new world of leading in this field."
Comment from Mr. Tomofumi Nakayama, Chief Product Officer, Director, Karakuri Co.
I am truly thrilled to launch 'AI for Science' with Professor Okada of the University of Tokyo and Upstage, a company that has been taking on challenges together. Many members of Karakuri's R&D team, including myself, are former members of the Okada Lab, and this project is a "return to our roots" for us.
We have also formed a strong team with Upstage to develop "Syn Pro," a top-class Japanese-language model in Japan. By connecting these dots, we hope to vigorously support researchers who are taking on the challenges of the future of science."
Comment by Hiroyuki Matsushita, President, Upstage AI K.K.
Since the kick-off of this project, Upstage and Karakuri have been exploring the possibilities of AI applications in the field of scientific research together for about six months. Upstage has been supporting the use of knowledge in research and business by transforming unstructured data into a form that AI can understand, based on its Document Intelligence technology cultivated in the global market. Upstage has been supporting knowledge utilization in research and business by transforming unstructured data into a form that AI can understand.
Furthermore, in October of this year, we released "SynPro," an LLM that boasts the highest level of Japanese-language performance as a model developed in Japan, in collaboration with Karakuri, and prepared a new foundation to realize data analysis and knowledge structuring in scientific research in a "safe and secure environment" and with "high cost efficiency.
By creating new synergies through the combined efforts of this project team, we hope to work together to create value at each of these sites as we look to a future in which AI will open up great possibilities for scientific research and industrial development in Japan and beyond."
Toward a future where AI becomes a "research partner
The "AI for Science" project presents a new approach that positions AI not as a mere computational tool, but as a scientific partner that assists researchers in their thinking and reasoning.
The project will utilize AI to increase the reproducibility and explainability of research and accelerate the generation of scientific knowledge, with plans to build prototypes and publish joint research results in 2026.
In the future, we aim to establish an ecosystem in which AI actually functions in the field of science, not only in the academic field, but also in collaboration with industry.
About Upstage
Upstage is a global AI startup that provides secure, high-performance generative AI solutions and is funded by Amazon and AMD.
Upstage has released a new next-generation LLM (Large Language Model) specialized for Japanese, providing highly accurate and reliable AI in the Japanese context.
Leaders in regulated industries such as insurance, legal, and healthcare are leveraging Upstage's AI technology to transform complex, unstructured documents into actionable business value.
| name of company | Upstage AI Inc. |
| Establishment | November 2024 |
| Location | 3F Mitsuhashi Building, 1-3-3 Kita-Aoyama, Minato-ku, Tokyo |
| representative | Hiroyuki Matsushita |
| uniform resouce locator | https://www.upstage.ai/ |









