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:20260611T145139Z
LOCATION:Bldg. 6 - Room 003
DTSTART;TZID=Europe/Stockholm:20260701T123000
DTEND;TZID=Europe/Stockholm:20260701T130000
UID:submissions.pasc-conference.org_PASC26_sess173_pap118@linklings.com
SUMMARY:CelloAI: Leveraging Large Language Models for HPC Software Develop
 ment in High Energy Physics
DESCRIPTION:Mohammad Atif, Kriti Chopra, Ozgur Kilic, Tianle Wang, and Zhi
 hua Dong (Brookhaven National Laboratory); Charles Leggett (Lawrence Berke
 ley National Laboratory (LBNL)); Meifeng Lin (Brookhaven National Laborato
 ry); Paolo Calafiura (Lawrence Berkeley National Laboratory (LBNL)); and S
 alman Habib (Argonne National Laboratory (ANL))\n\nNext-generation High En
 ergy Physics (HEP) experiments will generate unprecedented data volumes, n
 ecessitating High Performance Computing (HPC) integration alongside tradit
 ional high-throughput computing. However, HPC adoption in HEP is hindered 
 by the challenge of porting legacy software to heterogeneous architectures
  and the sparse documentation of these complex scientific codebases. We pr
 esent CelloAI, a locally hosted coding assistant that leverages Large Lang
 uage Models with Retrieval-Augmented Generation to support High Energy Phy
 sics code documentation and generation. This local deployment ensures data
  privacy, eliminates recurring costs, and provides access to large context
  windows without external dependencies. CelloAI addresses code documentati
 on and code generation through specialized components. For code documentat
 ion, the assistant provides: (a) Doxygen style comment generation by retri
 eving relevant information from text sources, (b) File-level summary gener
 ation, and (c) An interactive chatbot for code comprehension queries. For 
 code generation, CelloAI employs syntax-aware chunking that preserve synta
 ctic boundaries during embedding thus improving retrieval accuracy in larg
 e codebases. The system integrates callgraph knowledge to maintain depende
 ncy awareness during code modifications and provides AI-generated suggesti
 ons for performance optimization and accurate refactoring. Our results dem
 onstrate that CelloAI can enhance code understanding and streamline certai
 n development workflows, however domain expert oversight and validation is
  critical for reliable use of LLM-assistants in scientific computing conte
 xts.\n\nDomain: Engineering, Life Sciences, Physics\n\nSession Chair: Andr
 ei Onut (University of Basel)\n\n
END:VEVENT
END:VCALENDAR