ABE BioComplexity Seminar- Dr. Nikolaos Tziolas

Details

Speaker:
Dr. Nikolaos Tziolas
Title:
Spectral intelligence and conversational AI platforms for soil monitoring
Bio:
Dr. Nikos Tziolas is an Assistant Professor in the Soil, Water and Ecosystem Sciences Department at the University of Florida, based at the Southwest Florida Research and Education Center. He leads the Soil Science Artificial Intelligence (AI) laboratory, where his research and extension program lies at the intersection of AI, remote sensing, and digital soil modeling, with a strong focus on agro-environmental monitoring. His work leverages multimodal Earth observation data, including satellite imagery, and in situ sensors, to model dynamic soil properties, and ecosystem services. Dr. Tziolas integrates edge computing frameworks for real-time, in-field data processing and develops scalable AI-powered toolkits, apps and conversational AI platforms in support of stakeholders. His group is exploring large language models to enhance access to geospatial products and environmental data through AI agents, while also developing foundation models that deliver adaptable and transferable environmental insights—bridging the gap between complex analytics and practical use.
Abstract:
Recent advances in artificial intelligence (AI) and the growing availability of multimodal Earth observation data are creating new opportunities for agro-environmental monitoring. However, key challenges remain—especially the limited scalability and accessibility of current AI models. Many existing solutions are highly specialized, tailored to narrow tasks like soil property estimation, and often depend on specific sensors. This narrow scope limits their transferability and broader use.
At the same time, the outputs generated by these AI systems are frequently difficult for non-experts to access or interpret. Despite the abundance of Earth observation data and AI-driven insights, barriers such as technical interfaces, and a lack of user-friendly tools continue to non-expert users from using this information effectively.
This presentation introduces two solutions to these challenges. First, we present a spectral foundation model that integrates data from multiple VNIR–SWIR sensors to estimate soil properties across different regions. Second, we introduce a conversational AI system that combines large language models with geospatial intelligence to deliver agronomic insights through an interactive chat interface. This system allows users to query satellite data and crop-soil maps using natural language, making it possible to access tools for soil health mapping, cost-effective sampling design, and land degradation analysis with ease.