Industry, Research & Users Symposium

IEEE HiPC 2026 IRUS Session Proposal Industry Research and User Symposium (IRUS) 

Title: Building Energy-Efficient Sovereign AI Infrastructure at Scale with Self-Reliant  Compute Intelligence Platforms

Session Focus: Sovereign AI Infrastructure & Localized Intelligence

• • Grid readiness, power, cooling, cost dynamics of national compute capacity and Computing as a  Service platform. 
  • Deployment of domain-specific, small, efficient foundation models vs. frontier-scale training. • Privacy-preserving architectures, data localization, and regulatory compliance constraints 

Executive Summary 

Sovereign AI is increasingly moving from national-scale compute ambitions toward production ready, energy-efficient infrastructure capable of supporting domain-specific and locally  governed intelligence. This evolution requires more than compute capacity. It depends on an  open and optimized software stack that can efficiently utilize diverse hardware platforms while  providing flexibility, portability, security, and long-term technology independence. 

This session brings together leading participants from the compute ecosystem to discuss  practical considerations across the sovereign AI stack, ranging from processor and accelerator  architectures to open-source AI frameworks, optimized math and runtime libraries, model  deployment platforms, and large-scale infrastructure operations. The discussion will examine  trade-offs between frontier-scale foundation models and smaller, efficient domain-specific  models, with a focus on inference efficiency, performance-per-watt, memory and networking  requirements, utilization, and total cost of ownership. 

The session will also explore how open-source frameworks and software ecosystems can  enable cross-architecture portability, reduce reliance on proprietary software dependencies,  and improve deployment flexibility across sovereign environments. Topics will include AI  training and inference optimization, heterogeneous execution, hardware-aware software tuning,  reusable open-source software components, privacy-preserving and confidential computing  architectures, data localization requirements, and regulatory compliance considerations. 

The objective is to provide enterprises, infrastructure providers, policymakers, and the HPC-AI  research community with practical insights into building scalable, cost-effective, energy efficient, and open sovereign AI infrastructure, where hardware innovation and open-source  software development work together to deliver sustainable and adaptable AI capabilities.

Scheduled Date 

Thursday, 17th December 2026

Presenter(s) / Session Lead(s) and Affiliation 

This proposed HiPC 2026 IRUS workshop brings together an exceptional panel of leaders from  Fujitsu, Arm, AWS, AMD, and IISc, offering a unique blend of industry, cloud, semiconductor,  systems architecture, and academic research perspectives on Sovereign AI. 

The sessions collectively span the complete stack, from next-generation energy-efficient  processors and CPU-centric AI infrastructure to cloud-scale deployment architectures, open  heterogeneous computing platforms, and AI-driven scientific simulation frameworks. With  representation from global technology providers and leading academic researchers, the  workshop will provide attendees with comprehensive insights into building scalable, secure,  and self-reliant AI infrastructure ecosystems. 

We are pleased to confirm that all invited speakers have accepted the invitation and  confirmed their participation, ensuring a highly engaging program followed by an expert panel  discussion covering both enterprise adoption and cutting-edge research directions for  Sovereign AI. 

Here is the speaker list 

S.no.	Organization	Presenter Name	Designation
1	Fujitsu	Dr. Priyanka Sharma	Executive Director & BU Head,  MONAKA Software R&D Fujitsu Research of India
2	ARM	Mr. Ashok Bhat	Senior Product Manager, ARM
3	AMD	Dr. Prakash Raghavendra	Senior Fellow, AMD
4	AWS	Dr. Manish Modani	Senior AI Specialist, AWS
5	IISc	Prof. Phani Motamarri	Associate Professor, Computational  and Data Science Department, IISc

 

Moderators: Ragesh Hajela & Laks Raghupathi 

Symposium Leadership from HiPC 2026 Organizing Team: 

• • Laks Raghupathi, IRUS Chair & IEEE HiPC Vice General Co-chair (Shell India)
  • Srinivas Telukunta, Session Chair (Faculty Mentor, Johns Hopkins University & Great  Learning) 
  • Chiranjib Sur, IEEE HiPC General Vice-chair (Shell and Krea University)
  • HiPC 2026 IRUS CFP Details - https://hipc.org/irus-2026/ 

Point of Contact: Ragesh Hajela, Sr. Engineering Manager, MONAKA Software R&D Lab,  Fujitsu Research of India Pvt. Ltd. (FRIPL)

Organization	Presenter Name	Designation
Fujitsu	Dr. Priyanka Sharma	Executive Director & BU Head, MONAKA  Software R&D, Fujitsu Research of India
Photo
Speaker Bio	Dr. Priyanka Sharma is a technology leader working as the Executive Director  of Fujitsu Research of India & Senior Director of Software Engineering, where  she leads the FUJITSU-MONAKA Software R&D BU.  She represents Fujitsu as a Founding Steering Committee Member of the  Unified Acceleration (UXL) Foundation, collaborating with global technology  leaders on open, democratized AI acceleration standards. She is also  Governing Body member of the IWA Indian National Science Academy (INSA).  With over 25 years of technology leadership spanning industry, academia,  and research, Priyanka specializes in AI-enabled system design,  development, and deployment across HPC and systems engineering. She  has led global research collaborations and AI advisory engagements across  India, the UK, Canada, the USA, and the UAE.  LinkedIn - https://www.linkedin.com/in/drpriyankasharma/
Talk Title	Architecting the future of Sovereign AI with Open-Source Software,  Energy-Efficient & Secure Computing
Talk Abstract	Sovereign AI requires inference infrastructure that is energy-efficient, secure,  and built on open software ecosystems. This talk will highlight Fujitsu’s  approach to optimizing AI inference on CPU platforms, including open-source  contributions for inference performance and efficiency, together with  confidential computing technologies that protect sensitive AI workloads. The  goal is to enable sustainable, trusted, and self-reliant AI infrastructure for  sovereign deployments.

 

Organization	Presenter Name	Designation
ARM	Mr. Ashok Bhat	Senior Product Manager, ARM
Photo
Speaker Bio	Ashok Bhat is Senior Product Manager at Arm’s server business unit. He is a  member of a global product management team responsible for cloud and  HPC software, including ML  LinkedIn - https://www.linkedin.com/in/ashoki/
Talk Title	Agentic AI at Scale: Energy-Efficient Computing for Sovereign AI
Talk Abstract	Agentic AI is driving new requirements for always-on, responsive and energy efficient compute infrastructure. This talk will explore the critical role of CPUs  and the Arm ecosystem in enabling agentic AI, from orchestration and  reasoning workflows to efficient inference across heterogeneous  environments. It will highlight how open software ecosystems, CPU efficiency  and heterogeneous computing can provide a scalable and sustainable  foundation for Sovereign Agentic AI.

 

Organization	Presenter Name	Designation
AMD	Dr. Prakash Raghavendra	Senior Fellow, AMD
Photo
Speaker Bio	Dr. Prakash Raghavendra is PhD from the Department of CSA, IISc, in  Parallelizing Compilers. He has worked in HPE and Adobe before joining AMD,  Bangalore in 2012. Since then, he has been working as Sr. Fellow in AMD,  Bangalore, in AI and server software optimizations. Dr. Prakash has several  International Conference/Journal papers and patents to his credit.  LinkedIn - https://www.linkedin.com/in/prakash-raghavendra-8586578/
Talk Title	Open heterogeneous compute for Sovereign AI at scale
Talk Abstract	Dr. Prakash will introduce elements of AI compute infrastructure which  makes it successful for Sovereign AI especially India and how AMD is helping  enable this

 

Organization	Presenter Name	Designation
IISc	Prof. Phani Motamarri	Associate Professor, Computational and Data  Science Department, IISc
Photo
Speaker Bio	Prof. Phani Motamarri is an Associate Professor in the Department of  Computational and Data Sciences at the Indian Institute of Science (IISc),  Bangalore.  He heads the MATRIX lab with his research focussing on developing  mathematical techniques and hardware-aware computational methods  bridging advances in high-performance computing, quantum computing,  artificial intelligence with applications to quantum modeling of materials.  His group has pioneered highly scalable, finite-element-based algorithms for  density functional theory (DFT) calculations on emerging computing  architectures. These developments underpin DFT-FE, the open-source  electronic-structure code, enabling DFT simulations of unprecedented scale  and complexity.  He has received several prestigious awards and recognitions, including the  Robert J. Melosh Medal (2014), ACM Gordon Bell Prize Finalist (2019), ACM  Gordon Bell Prize (2023), Google India Research Award (2023), Dr. A.P.J.  Abdul Kalam HPC Award (2025), and Prof. Preeti Shankar Teaching  Excellence Award (2026).  LinkedIn - https://www.linkedin.com/in/phani-motamarri-013093a/
Talk Title	Towards Energy-Efficient Computing for Next-Generation Discoveries:  Low-Precision Hardware-Aware Algorithms meet Scientific ML
Talk Abstract	Scientific discovery increasingly depends on large-scale computations,  which must be made far more energy-efficient to remain sustainable. To this end, the talk presents a unified path on two fronts. First, we develop computational algorithms co-designed with modern CPU  and GPU hardware, exploiting low-precision arithmetic and lossy MPI  compression, in conjunction with matrix-free methods. These significantly reduce the time-to-solution while preserving accuracy, thereby enabling high-fidelity simulations at exascale.  Second, we design active-learning AI loops that couple these high-fidelity,  large-scale simulations with scientific machine learning strategies to build predictive models from far fewer data points labeled by these simulations.  Together, we demonstrate that these advances accelerate next-generation materials discovery while sharply reducing compute and data demands,  pointing toward sustainable, self-reliant AI-for-science infrastructure.

 

Organization	Presenter Name	Designation
AWS	Dr. Manish Modani	Senior AI Specialist, AWS
Photo
Speaker Bio	Dr. Manish Modani is currently working as Senior AI Specialist at AWS.  Previously he has worked as principal solution architect at NVIDIA. He has  served as technical lead for the National Supercomputing Mission, National  Monsoon Mission, India AI Mission, and National Quantum Mission. Manish  has successfully ported and benchmarked applications across Power, x86,  ARM, and PARAM architectures. His research is patented and published in  peer-reviewed journals. During his PhD at IIT Delhi, he studied the dispersion  of air pollutants under low-wind conditions in India.  LinkedIn - https://www.linkedin.com/in/drmanishmodani/
Talk Title	Sovereign AI for Science: Accelerating Scientific Discovery at Scale
Talk Abstract	AI for Science is transforming scientific discovery by enabling researchers to  explore increasingly complex problems at scale. This talk will discuss how  scalable and sovereign AI infrastructure can accelerate scientific discovery  while addressing requirements around data sovereignty, security, efficiency,  and flexibility. It will highlight how cloud and AI technologies can help build  trusted, scalable environments for the next generation of AI-driven scientific  research and innovation.

 

CALL FOR PROPOSALS (CFP) | IEEE HiPC 2026

Industry Research and User Symposium (IRUS) 

📍 Location: Bangalore, India 

📅 Dates: December 17–19, 2026

The Industry Research and User Symposium (IRUS) is the premier global forum bridging the gap between platform providers and enterprise end-users of advanced compute, data, and AI systems. IRUS is neither a product showcase nor a traditional academic paper session. Instead, it serves as a collaborative platform where industry leaders, researchers, service providers, and governing bodies convene to brainstorm how emerging technologies can shape the future—and how we, as a community, can strategically drive that transformation. 

Past editions  (2025, 2024, 2022, 2021,  2019, 2018, 2017, 2016)  have featured industry leaders like NVIDIA, Amazon, Accenture, Google, Microsoft, and Shell, alongside disruptive startups and world-class academic institutions such as IISc, CERN, and TIFR. The insights and discussions from these sessions have played a pivotal role in evolving both the conference and the broader HPC/AI ecosystem.

While previous editions explored technological possibilities, the 2026 symposium focuses on the critical infrastructure, systems engineering, and economic realities required to operate autonomous and frontier AI systems reliably, affordably, and accountably at industrial scale.

🌐 2026 IRUS Theme:

Scaling the Frontier of AI and Compute: Sovereignty, Physical Autonomy, and Trust

The Narrative: The transition from speculative AI models to resilient, real-world deployment is the defining engineering challenge of our era. IRUS 2026 unites the foundational rigor of academic research with the scale and urgency of enterprise execution.

Whether you are an industry practitioner managing the unit economics of multi-agent workflows, a researcher navigating sovereign compute constraints, or an engineer building physical AI systems where failure is not an option—this symposium bridges the gap between theory and production. Join us to explore the interplay between imported silicon and sovereign infrastructure, the shift toward embodied autonomy, and the system architectures required to anchor non-deterministic platforms in absolute trust.

🔍 Broad Tracks for Speakers & Session Leads:

🤖 1. Multi-Agent Systems & Enterprise Scale

• Production reliability, durable execution, and observability of non-deterministic, long-horizon workflows.
• Unit economics, token optimization, and HPC scheduling for heterogeneous GPU estates.
• Knowledge graphs, enterprise ontologies, and RAG evaluation beyond single-run accuracy

🌍 2. Sovereign AI Infrastructure & Localized Intelligence

• Grid readiness, power, cooling, and cost dynamics of national compute capacity and GPUaaS platforms.
• Deployment of domain-specific, small, efficient foundation models vs. frontier-scale training.
• Privacy-preserving architectures, data localization, and regulatory compliance constraints

⚙️ 3. Physical AI & Industrial Embodied Autonomy

• Vision-Language-Action (VLA) models, robot learning, and high-fidelity sim-to-real transfer.
• Synthetic data generation and world models for inspection, manipulation, and autonomous mobility systems.
• Edge inference under strict power/thermal constraints and safety certifications in heavy industry, energy, and agriculture

🎯 Submit Your Proposal

We strongly encourage submissions supported by live deployments, operational data, and concrete evidence of business or technical impact over conceptual demonstrations. 

📩 Please send your expression of interest to [email protected] with:

1. Title of the talk or proposed session focus
2. Presenter(s) / Session Lead(s) and Affiliation
3. Executive Summary (Max 250 words, focusing on production metrics, architectural lessons, or deployment data)

Symposium Leadership:

• Srinivas Telukunta, Session Chair (Faculty Mentor, Johns Hopkins University & Great Learning)
• Laks Raghupathi, IRUS Chair & IEEE HiPC Vice General Co-chair (Shell India)
• Chiranjib Sur, IEEE HiPC General Vice-chair (Shell and Krea University)