Sorting tickets nobody wants to sort
A classifier sorts incoming Jira tickets, a person confirms, and every correction makes the next round better.
The problem
Bug and backlog tickets arrive in large numbers and have to be sorted by hand before anyone can act on them. It is slow work, and the categories drift depending on who does it.
What I built
A classification pipeline that suggests a category for every ticket, with a person in the loop who confirms or corrects it. Corrections flow back as training data. A dashboard shows where tickets pile up.
What came out
Built at work for a client. The data stays with the client, so this page stays short.
Built with
Python, scikit-learn, LLMs, a dashboard