Equipment Maintenance Cloud 'EMLink' Integrates Uniquely Fine-Tuned Multimodal AI into Mobile — First in Japan for Equipment Maintenance, Inspection Tasks Completed by Voice Alone

Equipment Maintenance Research Institute Co., Ltd., which provides the 'EMLink' cloud service for factory and infrastructure equipment maintenance, has integrated its uniquely fine-tuned multimodal AI (local LLM) into its mobile app. This is the first time in Japan (according to our research) that a mobile-oriented local LLM has been implemented in the equipment maintenance domain. In addition to natural language dialogue, all voice and image inputs can be processed directly on smartphones and tablets, allowing all on-site inspection tasks (checklist creation) to be completed solely by voice input.

A live demonstration of this feature will be available at our booth (K46-17) during 'Factory Innovation Week Osaka 2026,' held at Intex Osaka from May 13 (Wednesday) to May 15 (Friday), 2026.

EMLinkMobile Multimodal AI (URL: https://www.youtube.com/shorts/OWsjcBuufDM)

*You can view a demo video of EMLink Mobile AI at the link above. URL: https://www.youtube.com/shorts/OWsjcBuufDM

EMLink Mobile AI Demo

Background of Development

In equipment maintenance sites, paper and Excel are still commonly used for recording inspection results. While the use of tablets and smartphones is progressing, site-specific constraints such as inputting while wearing gloves, voice recognition accuracy in noisy environments, and communication limitations inside plants where radio waves do not reach, have hindered widespread adoption. To address these challenges, we have equipped the 'EMLink' mobile app with our uniquely fine-tuned multimodal AI (local LLM), designed to process voice and images directly on smartphones and tablets.

Main Features

Multimodal Support — High-speed processing of both voice and images directly on smartphones/tablets We have integrated our uniquely fine-tuned multimodal AI (local LLM) into a dedicated smartphone/tablet app. Both voice input from the microphone and images captured by the camera are interpreted and structured by the AI on the smartphone/tablet.

Leveraging the responsiveness unique to local LLMs, voice input can be performed stress-free.

On-site inspection tasks completed by voice alone By inputting voice into a conversational chat UI, an appropriate checklist is activated, and checklist items can be further entered by voice. For example, if you respond with '異常あり' (abnormal), the content is automatically mapped to each field in the inspection form, completing a draft without any text input. Long inputs for supplementary information can also be handled seamlessly.

Support for field vocabulary through unique fine-tuning Generic AIs often fail to correctly recognize vocabulary used in factories and plant sites, such as 'part numbers,' 'process names,' and 'equipment nicknames.' Our AI is fine-tuned with our proprietary domain data to handle these terms. It corrects voice recognition text contextually and automatically converts it into structured text for inspection forms.

Offline operation The AI model itself, domain knowledge, and necessary data are all stored within the smartphone or tablet, allowing all functions, including voice and image processing, to be completed on the device. It can be used without functional limitations even inside factories where radio waves do not reach or in mountainous areas.

Company Information

Company Name: Equipment Maintenance Research Institute Co., Ltd. Established: September 2022 Representative: Shogo Aihara, Representative Director CEO Headquarters: 3-18-16 Megurohoncho, Meguro-ku, Tokyo Business Activities: Development and sales of asset management system 'EMLink' for factories and infrastructure, and related consulting. Website: https://em-labo.co.jp/ Major Shareholders: Meister Engineering Co., Ltd., DNX Ventures, Delight Ventures Keywords: