An Agentic AI Pipeline for an Evidence-Grounded Environmental Health Public Information System: Development and Evaluation

An agentic AI pipeline for the Salt Lake Valley integrates neighborhood-scale weather, air-quality observations, and asthma-related evidence to answer lay health-environment questions in one conversational interface.

• Why it matters: Users in urban areas with complex conditions struggle to combine heterogeneous environmental and health information into timely, actionable guidance, limiting translation into decisions.
• What they did: Built with LangGraph and OpenAI models, the system interprets natural-language queries, selects and runs external tools, retrieves real-time environmental data plus RAG asthma guidance, and synthesizes evidence-grounded responses with fallback robustness.
• The result: Benchmark evaluation showed generally strong tool recognition, parameter mapping, and incorporation of retrieved outputs across question types, supporting a scalable T3-to-T4 exposure health informatics translational step.

AI summary of the abstract · may contain errors · not medical advice · read the paper.