BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260421T090515Z
LOCATION:Bldg. 6 - Room 004
DTSTART;TZID=Europe/Stockholm:20260629T173000
DTEND;TZID=Europe/Stockholm:20260629T180000
UID:submissions.pasc-conference.org_PASC26_sess159_msa219@linklings.com
SUMMARY:UNIBE's GPUStack as an Example: Reporting on User needs
DESCRIPTION:Tobias Hodel (University of Bern)\n\nThe University of Bern ha
 s developed a proof-of-concept platform using GPUstack technology, providi
 ng a secure, institutionally controlled environment for deploying generati
 ve AI models. Designed for researchers working outside commercial cloud in
 frastructures, it enables direct management of computational pipelines on 
 self-hosted GPU resources, ensuring that sensitive data -- such as patient
  records, archival materials, or proprietary corpora -- remains within tru
 sted boundaries.\nThis approach addresses growing concerns around privacy,
  compliance, and data sovereignty. In healthcare, it supports model develo
 pment on protected patient data without external exposure. In the Digital 
 Humanities, it allows analysis of rare or restricted collections while res
 pecting custodial constraints. More broadly, it offers a framework for sec
 ure experimentation and reproducibility in NLP tasks involving confidentia
 l or regulated datasets. By integrating fine-tuned models with controlled 
 pipelines, the platform reduces reliance on third-party services while mai
 ntaining strong performance.\nIts flexible architecture supports users wit
 h varying expertise, improving accessibility without compromising security
 . The initiative also serves as a starting point for broader discussion on
  institutional AI infrastructure. Alongside introducing GPUstack, the proj
 ect presents survey results -- both internal and external -- highlighting 
 researchers’ needs and expectations for university-provided inference serv
 ices.\n\nDomain: Chemistry and Materials, Climate, Weather, and Earth Scie
 nces, Applied Social Sciences and Humanities, Engineering, Life Sciences, 
 Physics, Computational Methods and Applied Mathematics\n\nSession Chairs: 
 Tobias Hodel (University of Bern, Switzerland) and Sukanya Nath (Universit
 y of Bern)\n\n
END:VEVENT
END:VCALENDAR