ChatGPT’s release in 2022 has changed the way we process big data – for good.
Large-language models (LLMs), like ChatGPT, are garnering attention and popularity for a wide range of uses – but one application still flying under the radar is how these AI models can be integrated into data-rich business processes.
Join me as we look at how we are using Retrieval-Augmented Generation (RAG) to optimise our vast document base, saving us significant time in exchange for higher accuracy, efficiency and productivity.
Harnessing the Capabilities of RAG
The digital era has seen a surge in company documents due to digital transformation, increased data generation, regulatory requirements, collaborative work tools, affordable cloud storage, and the rise of Enterprise Content Management systems. AWL is over 60 years old and has a long history of working with technical documentation and codification, adding even more documents into the mix.
This begs the question: how can we efficiently find and distil information across this huge document base?
One common way to solve this using LLMs is RAG, which seamlessly combines the capabilities of LLMs with information retrieval. A RAG system retrieves the most relevant sections from the company’s documents. Then uses this text to generate outputs such as summarisations. This process ensures you are getting the most out of your LLM by ensuring they are informed by company data.
A RAG system offers transformative advantages for businesses. It allows for swift location and distillation of essential company knowledge, streamlining data access. Employees can quickly reference the company handbook for HR-related queries. I can facilitate efficient internal document reviews and research.
Envision an AI assistant backed by RAG, ready to provide insights based on your own sensitive data at just the click of a button.
We’ve just finished a prototype RAG system at Allan Webb and are now developing it further so it can be released to the rest of the company.
The emergence of LLMs like GPT-4 has revolutionised the way we approach many tasks, especially in data-rich environments like AWL. By utilising these tools, we’re enhancing our efficiency in handling the information in our vast document base.
We’re also exploring and researching how we can best leverage these new technologies in the defence sector, which I will be writing about soon. Watch this space!
I have a PhD in Physics with 5 years of data science industry experience. I have been the Lead Data Scientist at Allan Webb Limited since June 2023, and a data science consultant for 3 years previously. I am enthusiastic about NLP, LLMs, predictive analysis, and most importantly, solving our clients’ problems with data in whichever way is most effective and efficient.
Interested in learning more? Or simply want to chat? DM me or email me at firstname.lastname@example.org.