ETRI-Knowledge Sharing Plaform

KOREAN
논문 검색
Type SCI
Year ~ Keyword

Detail

Conference Paper Design and Implementation of an Agent-Based Multimodal MLOps Service Framework for AI Transformation in Manufacturing
Cited - time in scopus Share share facebook twitter linkedin kakaostory
Authors
HongYeon Yu, Jeongeun Kim, Aram Lee, Dong Hoon Son, Gihyeon Min, Seihyoung Lee
Issue Date
2026-10
Citation
International Conference on Information and Communication Technology Convergence (ICTC) 2026, pp.1-3
Publisher
IEEE
Language
English
Type
Conference Paper
Abstract
This paper proposes an agent-based MLOps service framework that enables manufacturing enterprise users to automatically develop AI models using the multimodal data they have collected. The proposed service framework automatically configures a suitable training task based on data validation and data profiling, and provides a service in which pre-trained open models are searched and evaluated to select candidate models optimized for the user's data, which are then fine-tuned with the user's data and deployed. The implementation results confirmed that the framework automatically constructs a task-compatible candidate set from user data, evaluates it under explicit policy criteria, and either selects a deployable model or refuses deployment with a recorded rationale.
Keyword
MLOps, multimodal models, agent orchestration, visual inspection, model selection agent
KSP Keywords
Candidate set, Data Validation, Data profiling, Service Framework, User data, Visual inspection, agent-based, design and implementation, manufacturing enterprise(ME), model selection, multimodal data