Prototyping an innovation assistant for Cleantech initiatives
2024-09-02 17:07
Today, most companies face some kind of environmental challenge - large companies are subject to tightening sustainability regulations, smaller companies suffer from increasing energy prices, and all of them need to resonate with customers who are increasingly aware of the limits of our planet. Cleantech provides solutions to many of these challenges - however, it is a deep-tech domain, so companies from other industries face huge knowledge gaps when considering Cleantech innovations.
Together with ETH Zurich and the universities of applied sciences in Lucerne and Northwestern Switzerland, we prototyped a conversational assistant to support companies with their Cleantech initiatives. The assistant relies on a dataset of Cleantech patents and media articles. Using a RAG architecture (Retrieval-Augmented Generation), it combines this specialised knowledge with the power of open-source LLMs such as Llama 3. Thus, it is able to answer specialised questions, suggest suitable innovations, and provide implementation guidance, supporting users on their Cleantech journey.
More than 40 graduate students participated in this challenge, and the results were presented at a workshop at the Swisstext 2024 conference. As practitioners know, the devil is in the details for RAG systems, so this experimental approach allowed us to test and evaluate different RAG approaches. During the coming months, we aim to provide our customers with an MVP of the innovation assistant.
If you are interested in supporting this development and would like to co-create an innovation assistant for your own domain of interest, get in touch at info@anacode.de and let’s talk!