This project introduces WikiBioM: a sophisticated AI system for the agricultural sector based on a hybrid approach of symbolic ontologies and fine-tuned Large Language Models. By bridging expert knowledge with generative AI, WikiBioM provides not only real-time information but also transparent, reasoned answers (explainability), offering critical decision support for agricultural management.

WikiBioM is designed to be scalable, catering to both small-scale farmers and industrial agricultural operations. Central to its design is the integration of sustainable and ecological farming principles, ensuring that generated recommendations minimize the use of pesticides and reduce negative impacts on the ecosystem.

Through the strategic implementation of AI, WikiBioM streamlines the flow of information and increases efficiency without compromising environmental integrity. Operating as a tailored 24/7 virtual assistant, the system supports a wide array of applications—including weather forecasting, plant care, pest management, and strategic harvest planning. Our goal is to drive the digital transformation of the industry, enabling farmers to leverage data-driven management to optimize productivity and cost-efficiency while fostering a more sustainable future.

Our approach bridges the gap between generative AI and structured expertise by combining Large Language Models with symbolic knowledge. We build a foundation of verified truth by synthesizing data from external streams—like weather and geospatial services—and internal specialized knowledge, including soil science ontologies and real-world insights gathered from farmers.

This refined data is then used to enhance the LLM through fine-tuning and Retrieval-Augmented Generation (RAG). By layering a formal semantic ontology over the AI, we move beyond simple text generation to deliver a system capable of providing high-precision, verifiable, and explainable insights to the user.