BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.adass.org//adass2026//speaker//RVNAQU
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-QSYMQC@pretalx.adass.org
DTSTART;TZID=AWST:20261103T152700
DTEND;TZID=AWST:20261103T153900
DESCRIPTION:Two options exist today\, and each falls short in a different d
 irection. Point a commercial AI application at data held on a science plat
 form and the science turns into plumbing: credentials\, staging\, transfer
 s\, all for a dataset that should never have moved. Run a coding CLI insid
 e the platform instead and it sits beside the data but sees only a shell a
 nd a filesystem — nothing of the archive\, the session lifecycle\, or th
 e access rules governing them. The first cannot reach the data\; the secon
 d cannot see the platform. Neither was built for research.\nThe scientific
  workbenches now appearing run on the researcher's own machine or cluster 
 so the data need not move. But their domain readiness is built for the lif
 e sciences\, and the integration path is proprietary: the platform arrives
  as a plug-in inside the vendor's application\, and research context accum
 ulates in that vendor's formats. Astronomy already has platforms tuned ove
 r decades to its data models\, access rules and standards\; we should not 
 have to migrate onto a general-purpose workbench to benefit from these mod
 els — the models should meet our platforms on our terms.\nI illustrate t
 his with a client over CANFAR\, Skaha and the CADC archives whose agent to
 ols are not a parallel API: each performs the same action as a control in 
 the interface\, so astronomers and agents share one surface. Both launch\,
  monitor and stop interactive and headless sessions\, and the agent can ta
 sk one with a script in an image carrying the astronomy libraries. Retriev
 ed observations collected in a research area\, products land in VOSpace\, 
 FITS and cube viewers give a fast visual read the model can be asked to as
 sess\, and workflows serve as written protocols the agent follows.\nAs thi
 s work is ongoing\, the exact final vision for the user-facing product is 
 still actively taking shape. Rather than presenting a finished system\, we
  present our integration across CANFAR\, Skaha\, and CADC as the foundatio
 n and core of a future AI-enabled science platform. We bring these develop
 ments to ADASS to open a dialogue with the community—seeking to learn ho
 w researchers envision the natural blend of AI models and research workflo
 ws\, and how our shared platforms can best deliver that vision.
DTSTAMP:20261116T153721Z
LOCATION:Seminar Room 2
SUMMARY:Bring Your Own Model: Treating AI as a Replaceable Part of the Scie
 nce Platform - Serhii Zautkin
URL:https://pretalx.adass.org/adass2026/talk/QSYMQC/
END:VEVENT
END:VCALENDAR
