(6 month intership + thesis)
Project Overview
It is widely acknowledged that LLMs can help save time in a wide range of work-related activities. However, their reliability is still a point of weakness and a subject of debate when discussing their application. In collaboration with the Lombardy office of the Court of Auditors, the work is underway to develop an application that harnesses the power of the best generative artificial intelligence models without compromising on accuracy, reliability and security. The aim of this thesis is to study the main techniques of LLMs hallucination mitigation or information extraction and apply them to a real-world use case. Over the course of the project, you will be required to familiarise yourself with textual data preparation, embedding representation, RAG and result evaluation techniques, with a view to creating a pipeline in which the LLM model used is optimised for its specific role.
Contacts
If you are interested, please send an email to g.lopiano1@campus.unimib.com, with davide.ciucci@unimib.it as cc to schedule a meeting.