ProjectFinderAgent

ETHGlobalProjectFinder is an AI-powered tool that enhances the way users explore ETHGlobal hackathon projects. While the original ETHGlobal website offers a basic search, our project builds on top of that by enabling a smarter, more intuitive search experience. Users can enter natural language queries to find relevant projects, and receive results that include the project’s name, a short description, and a direct link. It’s designed to make discovering past ETHGlobal projects easier, faster, and more relevant for developers, researchers, and enthusiasts.

We began by scraping publicly available project data from the ETHGlobal website, including names, descriptions, and links. This data was processed using an embedding model and stored in a vector database to enable semantic search. A uAgent serves as the interface between the user and the backend, forwarding queries to the ASI:ONE LLM. The LLM interprets the user's intent, retrieves the most relevant project entries from the vector store, and returns a refined list that includes each project’s title, summary, and link — delivering an enhanced and intelligent search experience.