Support assistant for Monroe-Woodbury CSD faculty, who sign in with Google SSO. A classifier trained on 10,000+ resolved tickets sorts each issue and suggests a fix, later extended with OpenAI's API.
Faculty at Monroe-Woodbury CSD reported technology problems through a help desk that handled a steady stream of tickets. I built a support chatbot that could answer the common ones right away.
At its core is an issue-classification model built with Brain.js and trained on more than 10,000 tickets that had already been resolved by people. Given a new problem, it predicts the category and returns the most relevant fix.
The app is full-stack: a Next.js frontend, a Node.js backend using WebSockets for real-time chat, MongoDB for storage, and Google OAuth so faculty sign in with their school Google accounts through single sign-on.
It was built before large language models were widely available. Later, I integrated OpenAI's API alongside the existing classification model.