Confidential

Type of project

Duration

Official website

The project at a glance

As part of a collaboration with a major international company in the electronics sector (with over 10,000 employees and €3 billion in revenue), CATIE was brought in to address the need to improve the ergonomics and accessibility of an application used by professionals in the field.

The objective was to make the application easier to use by designing an interface driven by natural language, allowing users to interact directly with the software through everyday language queries. CATIE leveraged its expertise in natural language processing (NLP), human-computer interaction, and large language models (LLMs) to meet this challenge.

Achievements

CATIE exceeded initial expectations by proposing features that were previously unknown to the client, such as automatic evaluation of model outputs to improve the robustness of the chatbot, as well as the ability to search and rephrase information from the client’s documents based on natural language questions, using Retrieval-Augmented Generation (RAG).

The code was easily adopted by the client, and RAG was identified as a high-value component that should be the first to be integrated into the application. Finally, various ideas involving generative AI were discussed and could lead to future collaborations.