Kereval is involved in a new collaborative innovation project called PrediObso, which aims to transform the maintenance of complex equipment in operational condition (MOC) in order to reduce obsolescence.

This project, led by OBSAM (OBSolescence and Analysis Management), brings its business expertise and unique Ostral database, the first collaborative multi-domain and multi-family database in France.

It involves a third player, the BARDE research laboratory at IRISA (Institute for Research in Computer Science and Random Systems). Its team is developing the algorithmic architecture, integrating deep learning and human-machine interaction.

Kereval’s Artificial Intelligence (AI) team guarantees the reliability and trustworthiness of the system through a machine learning lifecycle management process, MLOps (Machine Learning Operations), and rigorous testing methodologies.

Innovation at the heart of the project

PrediObso does not just analyze internal data; it uses cutting-edge AI technologies such as LLM (Large Language Models) and RAG (Retrieval-Augmented Generation) to enrich its predictions from external sources and offer a conversational interface to aid decision-making.

Obsolescence is a critical issue for industries with high technological and sovereignty stakes. To address this, the PrediObso project aims to develop a sovereign AI model capable of assessing the sensitivity of industrial components to obsolescence and therefore to supply disruption.

This project, certified by the ID4MOBILITY and Images & Réseaux clusters and supported by the Brittany Region, strengthens French technological sovereignty by offering a solution that can be deployed in a secure environment.