• ai
• artificial intelligence
• data science
• incident
• machine learning
• ticket

Incident management has always been a complex and delicate task, requiring
the swift and accurate routing of issues to the appropriate resolver groups.
This is particularly crucial in large organizations, such as the one examined in
this thesis, and even more so when dealing with strategic data-related issues.
Misclassification or delays can disrupt interconnected processes, lead to information loss, and cause missed opportunities, ultimately affecting profitability
and credibility.
In this thesis, we introduce SPHINX as a framework designed to enhance
the company’s incident management process. We will briefly present the company, explain how our framework works, and highlight its benefits compared
to the current system. Additionally, we will delve into the technical complexities of handling such a task, discussing why certain models were preferred over
others. Finally, we will explore the integration of a more innovative approach
leveraging Retrieval-Augmented Generation (RAG) to extend the current implementation, enabling it to handle edge cases more effectively. Through this
work, we aim to demonstrate how machine learning can drive faster, more
accurate, and cost-effective incident resolution.