Mahdi Farnaghi
I am an Assistant Professor at the University of Twente, Faculty of Geo-Information Science and Earth Observation (ITC), Department of Geo-Information Processing (GIP), Enschede, The Netherlands.
My primary research interest has been in the intersection of artificial intelligence (AI) and geodata processing (or GeoAI, in short). During Ph.D., I worked on solutions based on semantic web concepts/technologies and AI to generate geodata processing workflows automatically. After the Ph.D., I extended my research further into the spatial data science and big data processing fields, where I have been using machine learning (ML) for spatiotemporal modelling in different application areas, including environmental monitoring, analysis and assessment, disaster management, and spatial epidemiology. I exploited big data processing technologies and deep learning (DL) methods to extract information from the massive amount of spatial data collected for environmental phenomena. I have also been working on geospatial social media analysis using Natural Language Processing (NLP) methods.
Currently, I am working in the following directions:
- Large Language Models (LLM) and Multimodal LLMs (MLLM)
- MLOps and automation of geoprocessing workflows
- eXplainable Artificial Intelligence (XAI)
- Spatial and spatiotemporal effects in modeling using ML/DL
- Spatial data fusion
Sessions
Modeling in Geographic Information Systems (GIS) is central to understanding the spatial and temporal dynamics of both natural and human-made processes. Yet, constructing such models often demands deep technical expertise and familiarity with domain-specific tools, posing a barrier for many GIS users. IntelliGeo addresses this challenge by integrating the power of Large Language Models (LLMs) directly into QGIS through an intuitive chat-based interface. This AI-powered assistant supports users in designing spatial processing models by interpreting natural language instructions and translating them into executable GIS workflows.
Built entirely on open-source technologies, IntelliGeo’s backend rely on general-purpose LLMs to facilitate interactive modeling within QGIS. Each user interaction helps build a curated dataset aimed at fine-tuning LLMs for geospatial tasks. In doing so, IntelliGeo serves a dual purpose: it empowers users to create complex GIS models more easily and contributes to the development of a domain-adapted LLM for Geographic Information Science.
In this talk, we will describe the core functionalities of the IntelliGeo plugin, outline the
structure and goals of the emerging fine-tuning dataset, and share our roadmap for training a specialized LLM for GIS modeling and analysis.