Artificial intelligence for OAJ Observatory operations: from concept to practice with OPTIMIA, a multi-agent AI operational brain
Abstract
Building on our previous work on the preliminary AI design for operations at the OAJ, presented at the previous edition of this symposium, we report on the deployment of OPTIMIA, an AI-driven "operational brain" for the observatory. OPTIMIA is a multi-agent AI system integrated with the Observatory Control System (GOCS) and operational tools, providing AI-assisted support across all GOCS operation types. Using on-premises large language models and retrieval augmented generation over telemetry, alarms, maintenance records, logs and procedures, OPTIMIA offers conversational and voice interfaces that help staff diagnose faults, prioritise incidents and execute complex procedures. Exploiting its ability to connect large volumes of heterogeneous data, OPTIMIA becomes a powerful guide for orientation, coordination and management. We also outline the risks and challenges of being early adopters of this technology and how careful planning is essential to gain benefits without compromising safety or reliability.