Fine tune LLM with industrial data

[RutulPatel8]

Hi,

I'm currently working on how best to implement AI functionality in our project. However, I’m a bit uncertain about the approach we should take.

Project Overview:
We have an ERP system that manages a wide range of tenant data related to customer purchasing information. We want to integrate an AI solution for the admin of each tenant, allowing them to ask questions about business expansion, such as sales figures for a specific month last year, product details, and other business-related queries.

From my research, it seems that handling these kinds of questions effectively might require a Large Language Model (LLM). However, I understand that developing an LLM from scratch is a complex task that involves building a model and training it with vast amounts of data.

Given this, we are considering using a pre-trained model, hosting it on our server, and fine-tuning it with our specific data. Does this seem like a sound approach, or do you have any alternative suggestions?

Additionally, could you recommend any pre-trained models that would be suitable for this type of application, or perhaps suggest other AI solutions that might be a better fit?