Conference Agenda

Urban agriculture (UA) represents a critical solution for food security in urban environments; however, its widespread adoption is limited by the lack of specialized and personalized guidance that adapts to residents' space constraints and microclimate conditions. This study presents GROWIA, a web application leveraging Claude Sonnet 4 Large Language Model (LLM) through specialized prompt engineering to democratize agricultural knowledge and facilitate the production of nutritious foods in Metropolitan Lima. Unlike traditional rule-based systems requiring manual rule encoding or fine-tuning approaches demanding significant computational resources, GROWIA employs a carefully crafted prompt engineering methodology that transforms a general-purpose LLM into a domain-specific agricultural advisor. Implemented with a three-tier architecture and specialized prompt engineering, the application processes contextual user information to generate personalized cultivation plans through natural language interaction. Dual validation included technical evaluation by 8 agronomist experts on 19 cultivation plans and usability testing with 14 end users without experience. Results demonstrate a Correct Recommendations Rate of 89.5% (p=0.032), a System Usability Scale score of 85.2, a 55.4% reduction in perceived barriers, and 71.4% effective adoption. GROWIA demonstrates that domain-specialized LLMs through prompt engineering can effectively democratize agricultural expertise, offering a scalable, replicable solution for addressing urban food security challenges across Latin American metropolitan contexts.