FastLabel Inc. (Headquarters: Shinjuku-Ku, Tokyo; President and CEO: Takeshi Suzuki; hereinafter "FastLabel") announces that its proposal, "Research and Development on Technologies for Structuring Tacit Knowledge Held by Manufacturing Workers and Verification of Their Industrial Applicability," has been selected as a commissioned project under the project "Research and Development Project for Strengthening Post-5G Information and Communication System Infrastructure / Research and Development on AI-Readying Manufacturing Data and Other Industrial Data (GENIAC)," implemented by Japan's Ministry of Economy, Trade and Industry (METI) and the New Energy and Industrial Technology Development Organization (NEDO). This marks the first public call for proposals under GENIAC specifically focused on AI-readiness technologies. Under this project, FastLabel will collaborate with a leading Japanese automotive OEM to develop and demonstrate a general-purpose technology that converts tacit knowledge held by skilled factory workers such as intuitive judgments based on experience and fine-grained motion adjustment. Specifically, the project aims to achieve the automatic detection of tacit knowledge by labeling work videos, skeletal motion data, and eye‑tracking data of skilled frontline workers, using Vision‑Language Models (VLMs)*3 and difference analysis technologies. Background With the ongoing digital transformation of the economy and industry, data has emerged as a strategic asset that underpins productivity gains and innovation. Meanwhile, the publicly available data on the web that has supported improvements in AI performance is being increasingly used for AI training, making real-world data held by companies and organizations increasingly important. In addition, in the AI-related policies formulated by the government, the development and utilization of high-quality data is also positioned as one of the key pillars. While manufacturing remains one of Japan’s core industries, the sector is facing a shortage of skilled workers due to the country’s aging population and declining birthrate. As a result, the tacit knowledge that skilled workers have accumulated over many years is at risk of being lost as they retire. To date, many AI-readiness initiatives in manufacturing have focused primarily on organizing production system data and applying technologies such as Retrieval-Augmented Generation (RAG) and AI agents to existing documentation, including technical drawings and operational manuals. However, tacit knowledge that exists only in the experience and behaviors of frontline workers and has never been digitized. Effective methods for data collection and structuring have yet to be established. Since its founding, FastLabel has provided comprehensive solutions to support data-centric AI development, including the construction of data infrastructure that enables the development and operation of AI systems. Leveraging the expertise and methodologies cultivated through projects with leading Japanese corporations, FastLabel has developed the capability to generate high-quality datasets in specialized domains that require deep domain knowledge and technical expertise. Through this project, we will contribute to addressing the societal challenge of advancing AI adoption in Japan's manufacturing industry by tackling the untapped field of making tacit knowledge AI-ready. Project Overview ProgramName: Research and Development Project for Strengthening Post-5G Information and Communication System Infrastructure / Research and Development on AI-Readying Manufacturing Data and Other Industrial Data (GENIAC) Project Theme: Research and Development on Technologies for Structuring Tacit Knowledge Held by Manufacturing Workers and Verification of Their Industrial Applicability Overview: As part of this project, FastLabel conducted interviews and on-site investigations with skilled workers, managers, and executives at multiple manufacturing companies. These activities revealed that tacit knowledge is a common challenge across the manufacturing industry. Based on these findings, FastLabel has organized the stages of data preparation for AI adoption into the following three phases. • AI-Ready 1.0: Development and utilization of structured production data, including time-series data collected through Supervisory Control and Data Acquisition (SCADA) systems and Bills of Materials (BOMs). • AI-Ready 2.0: OCR processing and conversion into Retrieval-Augmented Generation (RAG)-ready data for existing unstructured data contained in technical drawings, operation manuals, and other documents. • AI-Ready 3.0: Extraction and accumulation of data on tacit knowledge that is difficult to collect and store as data. Under this project, with AI-Ready 3.0 as the goal, FastLabel will conduct the following four research and development initiatives to establish a reproducible process for extracting tacit knowledge that can be applied across companies in the same industry as well as in other industries. ① Multimodal Data Collection Methods and Technologies ② Methods and Technologies for Tacit Knowledge Structuring and Extraction ③ Evaluation Methods and Technologies for AI Models, Including Vision Language Models (VLMs) ④ Data Collection Support Systems Call for Proposals Details:https://www.nedo.go.jp/koubo/CD2_100422.htmlSelection Results:https://www.nedo.go.jp/koubo/CD3_100422.htmlFuture Outlook This project involves the end-to-end implementation of real-world data collection, data structuring, VLM training, and evaluation to make the tacit knowledge of skilled workers in the manufacturing industry AI-ready. We believe that the methods developed through this process could become essential technologies for the future development and evaluation of multimodal foundation models. In particular, the development of multimodal foundation models required for Physical AI requires large quantities of high-quality multimodal data. Establishing methods for collecting first-person perspective data from skilled workers and lowering the difficulty of data acquisition will contribute to development based on AI scaling laws, in which AI performance improves in proportion to the amount of data and other factors. In April 2026, FastLabel established its Robotics AI Business Division and has since been advancing proof-of-concept (PoC) research on Physical AI in collaboration with manufacturing companies. Leveraging the technologies developed through this project, FastLabel will not only promote the AI-readiness of tacit knowledge in the manufacturing industry but also establish an end-to-end solution covering the acquisition and structuring of first-person perspective data for use in Physical AI development projects. Furthermore, by sharing the expertise and technical knowledge gained through this project with Japan's Ministry of Economy, Trade and Industry (METI) and collaborating closely with the ministry, FastLabel will contribute to the growth of Physical AI in Japan—particularly AI robotics—from the perspective of data. CommentsTakuya Watanabe Director, AI Industry Strategy Office, Information Technology Industry Division, Commerce and Information Policy Bureau, Ministry of Economy, Trade and Industry (METI) This project aims to enhance real-world data held by companies and organizations into a state that enables AI utilization (AI-Ready), while strengthening the competitiveness of the industry as a whole through the demonstration of its effectiveness and the sharing of its outcomes. It is an important initiative for securing Japan's industrial competitiveness. In particular, FastLabel's initiative is comprehensive, encompassing the structuring of tacit knowledge from the factory floor, and we have high expectations for its outcomes. Takeshi Suzuki President and CEO, FastLabel Inc. We are honored that FastLabel's proposal has been selected in the first public call for proposals under the "Research and Development on AI-Readying Manufacturing Data and Other Industrial Data (GENIAC)" program, implemented by Japan's Ministry of Economy, Trade and Industry (METI) and the New Energy and Industrial Technology Development Organization (NEDO). As demonstrated by the scaling laws, AI performance improves in proportion to the amount of data available for training. Now that AI training on publicly available web data is approaching saturation, the next frontier lies in how to make the real-world data held across industries AI-ready. In particular, the tacit knowledge possessed by skilled workers in Japan's manufacturing sector is gradually being lost due to the country's aging population and declining birthrate. At the same time, we believe that structuring this knowledge and passing it on to future generations is a critical challenge that will shape Japan's competitiveness in the era of Physical AI. Since its founding, FastLabel has been building the data infrastructure that supports data-centric AI development and has assisted numerous manufacturing companies with their AI initiatives. Through this project, we will collaborate with a leading Japanese automotive OEM to take on the challenge of making tacit knowledge AI-ready. By applying the technologies and expertise established through this project across the manufacturing industry and to Physical AI development more broadly, we aim to contribute to the future of Japanese manufacturing and the advancement of the AI industry. Notes *1 AI Readiness AI-readiness refers to preparing data for AI utilization by making it well-structured (structuring and modeling), appropriately sized (chunking), semantically enriched (vectorization and labeling), high quality (with minimal errors and bias), consistently managed (governance and security), and continuously improved (monitoring and feedback). *2: GENIAC (Generative AI Accelerator Challenge) GENIAC is a project implemented by Japan's Ministry of Economy, Trade and Industry (METI) and the New Energy and Industrial Technology Development Organization (NEDO) to strengthen Japan's capabilities in generative AI development. The project provides computational resources for the development of foundation models, which are the core technology underpinning generative AI, and supports demonstration studies aimed at the utilization of data and AI. This is the first public call for proposals under the "Research and Development on AI-Readying Manufacturing Data and Other Industrial Data" program. ※3：VLM（Vision-Language Model） Vision Language Model (VLM) is a multimodal AI model that can simultaneously understand and process visual information, such as images and videos, and language information, such as text. About FastLabel FastLabel, Inc. is committed to building data infrastructure that enables data-centric AI development. The company provides end-to-end support across the AI development lifecycle, including data collection and generation, annotation, model development, and DataOps implementation. In recent years, FastLabel has expanded its focus to the field of physical AI, including robotics, and has been developing data pipelines that support the creation of robot foundation models and Vision-Language-Action (VLA) models. Company Overview Company Name: FastLabel Inc. Headquarters: Shinjuku Sumitomo Building 24F, 2-6-1 Nishi-Shinjuku, Shinjuku-ku, Tokyo, Japan Representative: Takeshi Suzuki, President and CEO Business: Provider of professional services and products supporting Data-centric AI development Website: https://fastlabel.ai/service/robotics