We’ve funded a range of projects to guide our roadmap for the future

A federated bespoke AI-assisted helpdesk for DRI facilities

DRI facilities, and in particular HPC systems are complex to use, with a steep learning curve for new users. Although similar in principle, each system is bespoke with its own design characteristics, meaning that knowledge gained on one system often cannot be directly transferred to others, adding confusion for users.

Helpdesks often spend significant time answering simple queries around login problems, or issues which are widely solved or for which information is available on the facility help pages.

This proposal seeks to investigate removing this burden from helpdesks by automating responses using bespoke AI Large Language Models (LLMs). The concept is simple: Retrieval Augmented Generation (RAG) techniques will be used to frame user queries and select the most relevant solution to the problem. The RAG will be trained on facility websites, and therefore highly relevant information will be returned.

This project will be key for smaller DRI systems which would not have the effort available to develop such a solution, and will free up the stretched technical support teams to concentrate on important issues.