[hpc-announce] CFP for ISAV: In Situ AI, Analysis and Visualization at #SC25

[Matthew Larsen]

https://urldefense.us/v3/__https://isav-workshop.github.io/2025/__;!!G_uCfscf7eWS!dseYGPCq8VZ8bIVYTuKmVNgcFAOjcS-pN7mi2Kzlo4i8-ut9Azll3qjWCnhtR_uaCAcOncKyeVTUR5eL0Cb0V0Fj$ 

In conjunction with: SC25, The International Conference for High
Performance Computing, Networking, Storage, and Analysis

Workshop Theme
------------------------
As HPC platforms and applications increase significantly in size,
complexity, and heterogeneity, one major challenge is the widening gap
between computation and our ability to gain insight from extreme-scale
data and make timely, data-driven decisions. A well-known, yet
challenging, approach is in situ processing – performing as much
analysis as possible while computed data is still resident in memory.

This is the 11th year of the ISAV workshop, and we are expanding the
workshop’s scope to In Situ AI, Analysis and Visualization.
Furthermore, we are expanding the workshop’s technical program to now
accept full paper submissions. We celebrate that in situ processing
has evolved from research efforts to a central component in
supercomputer, cloud and edge applications. In situ methods are in
high demand:

in training ML/AI surrogate models or leveraging ML/AI models for analysis,
in system-scale 3D visualization for the latest Exascale supercomputers,
in cloud products providing responsive user experiences,
in tightly coupled digital twins, and in computational sciences.
Each one of these examples has a different set of requirements in
response time, data throughput and complexity of data pipelines, and
more exploration in the in situ space is needed to address
multifaceted goals: (1) to preserve important elements of simulations,
(2) to significantly reduce the data needed to preserve these
elements, (3) to offer as much flexibility as possible for
post-processing exploration, and (4) to accelerate the gathering of
insights to be fast enough to make timely decisions based on it.

ISAV’25 brings together the broader HPC community and researchers,
developers and practitioners from industry, academia, and government
laboratories who are developing, applying, and deploying scalable in
situ methods at any high performance platform. The goal is to present
research findings, lessons learned, and insights related to developing
and applying in situ methods across a range of science and engineering
applications in scalable environments; to discuss topics like
opportunities presented by new workflows in AI/ML, modeling, data
processing, emerging architectures, infrastructure needs,
requirements, and gaps, and experiences to foster and enable in situ
AI, analysis, and visualization. Since its inception in 2015, ISAV has
fostered and catered a diverse audience and supported early career
members, becoming a “center of gravity” for researchers,
practitioners, and users/consumers of in situ methods, software and
integrations in the HPC space. Through presentations and discussions
of research findings, lessons learned, and early ideas, ISAV
illuminates new requirements and gaps driven by science and
engineering applications, and fosters the community members and
knowledge base around the development and application of in situ
methods with its peer-reviewed proceedings.

Participation/Call for Papers
--------------------------------------
In its 11th year ISAV is expanding in scope and technical focus, and
now invites full paper submissions up to 10 pages (including
references) and works on in situ AI/ML training or inference. ISAV
also continues to invite short papers (5 page + 1 page references) and
lightning talk abstracts (1 page).

Full papers should present research results, identity opportunities or
challenges, or present case studies/best practices for in situ
methods. Short papers may also document late breaking ideas & early
progress on novel concepts. Lightning talks are encouraged to present
preliminary works or ideas to foster discussion with the community.
Full and short papers will appear in the workshop proceedings and
authors will be invited to give an oral presentation at the workshop;
lightning talks will be invited to give brief oral presentations at
the workshop.

Submissions of all types may identify opportunities, challenges and
best practices for in situ AI/ML, in situ analysis and in situ
visualization. They may propose new methods and techniques, provide
positions, or experience reports on in situ analysis, learning and
visualization. Areas of interest for ISAV include, but are not limited
to:

Methods, Algorithms and Synthesis between HPC & ML: In situ analysis
(feature detection, data reduction/compression, data summarization, ML
training) and scientific visualization using data-driven,
surrogate-assisted, statistical, temporal, geometric, or time-varying
methods.
Applications and Workflows: Applications (simulations, data
processing, scientific user facilities) and integrations into digital
twins. Workflows for supporting complex in situ processing pipelines
(incl. enabling accelerated post-processing and elasticity), their
resilience (error detection, data congestion, fault recovery) and
reproducibility.
Scalability Requirements: Scalability, resource utilization, data
flow, and simplified access to extreme heterogeneous resources.
Real-time coupling of data (modeled or measured), surrogates and
algorithms.
Case Studies, Data Sources and Best Practices & Usability:
Examples/case studies of solving a specific science challenge with in
situ methods/infrastructure. In situ methods/systems applied to data
from simulations, and/or observations/experiments. Deployments &
software engineering.
Software Evolution & Standardization: In situ libraries from research
prototypes to production quality. Challenges, opportunities, gaps in
existing capabilities. API designs and development of community
standards.
Enabling Hardware & Emerging Architectures: Hardware & emerging system
architectures that provide opportunities for in situ processing.
Efficient use of hardware accelerators and heterogeneous
architectures, incl. HPC, Data Center or Edge.