Presentation Information
[18a-PA1-22]Development of a Generative AI Agent for Advanced Utilization of ARIM Shared Data: Supporting ALD Experimental Condition Selection Using the MCP and Knowledge Graphs
〇Takeshi Kuwada1, Hiroshi Arimoto2, Masashi Yamazaki2, Atsuko Onizawa2 (1.NIMS, 2.AIST)
Keywords:
Atomic Layer Deposition (ALD),Materials Informatics,Large Language Models (LLM)
We developed a generative AI agent using MCP and a knowledge graph for ALD experimental data published as shared data under the ARIM project. By linking sparse data measured under diverse conditions, the agent provides insights useful for selecting experimental parameters. In response to natural language queries, it indicates the availability of relevant data and suggests reference experiments or equipment records under similar conditions, thereby streamlining initial planning and promoting reuse of past data.