Samuel Vander Velpen (Victoria University of Wellington, Nelson AI) is discuss the development of an AI data collection tool for the GuardIAS project at the fourth annual Artificial Intelligence Researchers Association conference at Victoria University of Wellington, 17-18th November.
Paper Abstract
A significant challenge in invasive species management is the difficulty of gathering and combining data from many different sources. For a critical issue like invasive species, information is often scattered, inconsistent, and hard for researchers to use effectively. To address this, we are developing a Python-based web tool that uses AI to automate this complex data integration task.
The system connects a simple frontend interface to a powerful backend. A user makes a request by selecting topics from a list of buttons (e.g., “reproduction rates”, “prevention strategies”) and entering a species’ Latin name. This action triggers a multi-agent AI process on the backend. The approach uses several AI components that work together to query other specialized AI tools and databases. The system then extracts the key details and formats the data into standardized, usable files like CSV and GeoTIFF.
This work focuses primarily on the backend architecture, detailing how the multiple AI agents collaborate to unify and enrich data. More broadly, it provides a demonstration of how the latest agentic AI methods can be applied to solve complex data synthesis challenges. The workflow, from initial query to final structured output, offers a model for creating a cohesive knowledge base not only for invasive species management but also for other domains where data is highly distributed and fragmented.
View the slides: GIAS AI conference
