Aitor Ibarra Ibaibarriaga
Aitor Ibarra Ibaibarriaga is a systems and software engineer, scientific software architect, and space mission specialist with more than 25 years of experience supporting ESA science missions, primarily at the XMM-Newton Science Operations Centre at ESAC.
His background combines high-energy astrophysics, scientific software engineering, mission operations, robotic observatories, and the development of AI-based tools for space science. Since 2003, he has contributed to several areas of the XMM-Newton ground segment, including the Science Analysis System, scientific algorithms, mission-planning tools, operational data handling, and interfaces with Flight Dynamics and mission operations teams.
He has also contributed to ESA initiatives including CESAR, the Test-Bed Telescope project for Planetary Defence, ESA Datalabs, and the definition of the NewAthena Science Ground Segment. More recently, he has promoted the development of XMMGPT, an agentic AI initiative designed to improve access to mission knowledge and support scientific and operational workflows.
Session
Large Language Models are often presented as conversational interfaces, but their greater value for mission operations may lie in their ability to coordinate trusted tools, technical knowledge, and scientific services.
XMMGPT is an Agentic AI prototype developed for the XMM-Newton Science Operations Centre that combines Retrieval-Augmented Generation over mission documentation with access to operational and scientific services through machine-readable interfaces.
Building on the XMMGPT prototype previously presented at ADASS, this contribution shifts the focus from an individual mission assistant to a reusable architectural pattern for Science Operations Centres. We demonstrate how an LLM can act as a mediator between astronomers and existing mission capabilities, selecting and invoking specialised services rather than attempting to reproduce their functionality. Current use cases include querying technical documentation, checking target visibility, exploring X-ray catalogues through natural-language-to-ADQL workflows, retrieving and generating scientific products, estimating pile-up and count-rate conversions, and accessing upper-limit and long-term light-curve services. These capabilities support an interactive proposal preparation workflow in which users can move from a scientific question to feasibility assessment and archive exploration through a single conversational interface.
The approach relies on constrained, well-tested, agent-ready services with explicit schemas, controlled parameters, provenance, and human oversight. This separation between a mission-independent AI orchestration layer and mission-specific knowledge and tools offers a practical route towards reuse across multiple missions and observatories, while retaining scientific traceability and operational control.