On-Premises Large Language Models: safeguarding privacy during classification and privacy violation detection.

The following thesis describes the feasibility study and implementation of a pilot Business Intelligence solution for a financial partner operating in the automotive industry. This operator aims to manage issue and bug tracking ticket in order to: i) route each issue to the correct technical area, ii) identify, if any, all the personal and vehicle related data entered in each ticket with the goal of tracking it and assisting management in handling it in compliance with the DORA (Digital Operational Resilience Act) regulations. It is central for the client to keep all data within the company. They wish to introduce LLM to automate different business processes and this is an opportunity to show its potential.