BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.adass.org//adass2026//talk//U3X8EN
BEGIN:VTIMEZONE
TZID:AWST
BEGIN:STANDARD
DTSTART:20000101T000000
RRULE:FREQ=YEARLY;BYMONTH=1;UNTIL=20051231T160000Z
TZNAME:AWST
TZOFFSETFROM:+0800
TZOFFSETTO:+0800
END:STANDARD
BEGIN:STANDARD
DTSTART:20070325T040000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:AWST
TZOFFSETFROM:+0900
TZOFFSETTO:+0800
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20071028T030000
RRULE:FREQ=YEARLY;BYDAY=4SU;BYMONTH=10;UNTIL=20081025T190000Z
TZNAME:AWDT
TZOFFSETFROM:+0800
TZOFFSETTO:+0900
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-adass2026-U3X8EN@pretalx.adass.org
DTSTART;TZID=AWST:20261102T100900
DTEND;TZID=AWST:20261102T101800
DESCRIPTION:Large Language Models are often presented as conversational int
 erfaces\, but their greater value for mission operations may lie in their 
 ability to coordinate trusted tools\, technical knowledge\, and scientific
  services. \n\nXMMGPT is an Agentic AI prototype developed for the XMM-New
 ton Science Operations Centre that combines Retrieval-Augmented Generation
  over mission documentation with access to operational and scientific serv
 ices through machine-readable interfaces.\n\nBuilding on the XMMGPT protot
 ype previously presented at ADASS\, this contribution shifts the focus fro
 m an individual mission assistant to a reusable architectural pattern for 
 Science Operations Centres. We demonstrate how an LLM can act as a mediato
 r between astronomers and existing mission capabilities\, selecting and in
 voking specialised services rather than attempting to reproduce their func
 tionality. Current use cases include querying technical documentation\, ch
 ecking target visibility\, exploring X-ray catalogues through natural-lang
 uage-to-ADQL workflows\, retrieving and generating scientific products\, e
 stimating pile-up and count-rate conversions\, and accessing upper-limit a
 nd long-term light-curve services. These capabilities support an interacti
 ve proposal preparation workflow in which users can move from a scientific
  question to feasibility assessment and archive exploration through a sing
 le conversational interface.\n\nThe approach relies on constrained\, well-
 tested\, agent-ready services with explicit schemas\, controlled parameter
 s\, provenance\, and human oversight. This separation between a mission-in
 dependent AI orchestration layer and mission-specific knowledge and tools 
 offers a practical route towards reuse across multiple missions and observ
 atories\, while retaining scientific traceability and operational control.
DTSTAMP:20261116T154908Z
LOCATION:Seminar Room 2
SUMMARY:Beyond XMMGPT: Agentic AI for Science Operations Centres\, from Pro
 posal Preparation to Catalogue Exploration - Aitor Ibarra Ibaibarriaga
URL:https://pretalx.adass.org/adass2026/talk/U3X8EN/
END:VEVENT
END:VCALENDAR
