Sherpa & Company Selected for 'GENIAC', a National Project Accelerating Generative AI Implementation

Sherpa & Company (Headquarters: Shinagawa-ku, Tokyo; CEO: Jun Sugimoto; hereinafter 'Sherpa') is pleased to announce its selection as a 4th-batch operator for the 'Post-5G Information and Communication System Infrastructure Enhancement R&D Project / Development of Competitive Generative AI Foundation Models (GENIAC)' (hereinafter 'the Project'), administered by the Ministry of Economy, Trade and Industry (METI) and the New Energy and Industrial Technology Development Organization (NEDO).

Overview of the Selected Project The Project is a subsidy program implemented by METI and NEDO for companies and universities developing competitive generative AI foundation models, which are a core technology for information and communication systems. Following a rigorous screening process, the company was selected for the 4th batch of this project.

The company proposed the theme of 'Constructing Unstructured Document Analysis and Evidence Generation Models for the Financial Sector' and will carry out the development and validation of the following two models. The company aims to construct a foundation model capable of extracting information with high precision and presenting the evidence for complex Japanese unstructured documents, which have been a significant challenge in financial practice.

1. Development of a Document Analysis Model The company will build a model to analyze and understand documents with complex layouts, such as integrated reports, securities reports, and financial explanation materials.

2. Development of an Evidence Generation Model The company will build a model that provides information based on document content in response to user queries and presents the supporting evidence.

This initiative aims to significantly reduce the man-hours required to verify the accuracy of extracted results and to improve the reliability and transparency of information in the financial sector. By developing a model that also handles Japanese-specific document formats using domestic infrastructure, the company will realize high-precision analysis capabilities for documents produced in Japan. The insights gained during development, the developed models, and evaluation datasets are planned to be partially disclosed, contributing to the enhancement of Japan's generative AI research and development capabilities.

NEDO Project Overview: https://www.nedo.go.jp/koubo/CD3_100421.html

Project Details Implementing Agencies: METI and NEDO Project Name: Development of Competitive Generative AI Foundation Models (GENIAC) Theme: Constructing Unstructured Document Analysis and Evidence Generation Models for the Financial Sector Period: One year starting from July 2026 to July 2027

Validation In the validation of this theme, the company will collaborate with major institutions that support financial market decision-making to conduct verification through non-financial research tasks. The goal is to confirm the effectiveness in streamlining information confirmation tasks and improving transparency for information collection tasks that previously required a significant amount of time.

Background of the Project The company has been promoting business support for institutional investors and financial institutions through the provision of 'SmartESG Clarity'. However, in practical settings, the analysis of unstructured documents such as corporate integrated reports remains an analog process dependent on human resources, where staff manually extract information and checkers perform final confirmation by comparing the evidence with the original text. In particular, the task of accurately reading over 1,000 items of information per company from documents with complex layouts and recording and checking individual evidence is a significant barrier to highly advanced data utilization due to the high workload.

Given this background, the company will leverage its expertise in the financial sector to begin development of document analysis and evidence generation models. By building a more reliable generative AI model based on its understanding of customer needs cultivated in its existing business, the company will lead the advancement of information distribution across Japan.

Comments from Jun Sugimoto, CEO of Sherpa & Company 'To realize a world where profit and sustainability converge, we provide a data platform that supports decision-making in financial markets, including companies and investors. Currently, the opacity and inefficiency in extracting and verifying vast amounts of unstructured documents are structural issues in the financial market.

This project is positioned as an important initiative to accelerate our materiality of 'providing a sustainability data platform' and to increase information transparency through the ethical use of AI. We are confident that because we know the practical challenges, we can build an infrastructure that ensures the reliability of information distribution.'