Introducing AgriLLM: An AI powered agricultural advisory service for the global south

This presentation introduces AgriLLM, an open-source, agriculture-specific large language model and AI assistant designed to improve access to agricultural knowledge and advisory services in low-resource settings. The project aims to bridge information gaps by providing multilingual, crop- and region-specific recommendations through a combination of fine-tuned AI models and retrieval-augmented generation (RAG) that draws on verified agricultural knowledge bases. Developed through a global partnership of more than 15 organizations, AgriLLM emphasizes data quality, local validation, farmer participation, and rigorous evaluation to improve accuracy, reduce hallucinations, and ensure contextually relevant advice. Early results indicate that the model outperforms general-purpose AI systems on agricultural tasks and supports more reliable, domain-specific responses. The presentation highlights AgriLLM’s potential to strengthen agricultural extension services at scale while fostering collaboration, innovation, and equitable access to agricultural intelligence worldwide. -Abstract generated by Copilot AI 2.2