Adapting Large Language Models to Specific Domains

Large Language Models (LLMs) such, as Chat-GPT, are currently changing all the ways in which people interact with and get help from computers. Since the current changes in the fields of natural language processing and AI in general are so quick, then many aspects of LLMs are not understood, So, adapting them to specific domains is just starting. This includes healthcare, security, business, education, and many others.

This project will look at a proof-of-concept application to the field of the candidate’s choice. The supervisor can also suggest some choices. Additional challenges such as privacy and safety of the underlying technologies can be explored on the way.

While technical skills such as Maths and programming are useful, the comprehensive understanding of the technologies involved is not crucial, since the most work will be done in-field: designing, testing and assessing a specific application of a LLM. We will work both with in-context learning (ICL), fine-tuning (training) and neural architecture changes if needed. We will use Archie-West Supercomputer Center, which will provide unique opportunity to using LLMs directly, without relying on public commercial APIs.