Location: Room M2 – Melati Jr 4011/4111 (Level 4)
Abstract: This track explores the intersection of High Performance Computing (HPC) and Artificial Intelligence (AI) for accelerating key industrial, national, and global applications. Leading experts will discuss advancements in architecture and systems, including next-generation GPUs and AI accelerators. Topics include architectural breakthroughs, performance optimization, energy-efficient designs, and real-world applications. Attendees will gain insights into software-hardware co-design, tools for accelerating AI/ML models, and emerging hardware architectures for accelerating large models such as LLMs. This track enables collaboration among researchers, engineers and various industry players to tackle various aspects of AI development.
Track Chair: Dr Rick Goh, Director, Computing and Intelligence, IHPC, A*STAR
[Invited Track]
Programme:
| Time | Session |
| 01:30pm – 01:45pm | Opening – Dr Rick Goh, Director, Computing and Intelligence, IHPC, A*STAR |
| 01:45pm – 02:00pm | AI for Science Projects at Riken AI for Science is becoming one of the core research activities at Riken, Japan’s premiere national lab for Science. Among the major efforts, TRIP-AGIS is the flagship project where we seek to establish generative AI models for Science across various disciplines such as biology and material science. The outcomes of the project will heavily affect the next generation HPC flagship project in Japan, FugakuNEXT, where we hope to achieve zettascale performance in AI capabilities and deep synergy with traditional HPC. – Prof Satoshi Matsuoka, Professor, R-CCS |
| 02:00pm – 02:15pm |
What’s Next in AI Starts Here This talk explores how AI augments human intelligence instead of being at odds. Real-world examples increasingly demonstrate how AI partners with humans, from medical diagnostics to artistic creation. We discuss AI’s role in boosting creativity, decision-making, and productivity across various fields; and, the concept of human-AI symbiosis that emphasises collaboration over competition. The key takeaway: “The future isn’t AI vs. humans—it’s AI + humans.” – Dr Ng Aik Beng, Senior Regional Manager, NVIDIA AI Technology Center |
| 02:15pm – 02:30pm | An HPC-AI Fusion Approach Towards Earth System Modeling on Next-Generation Supercomputers The Earth, as one of the most enduring yet complex research subjects, remains a paramount target for computational modeling across the globe. Despite decades of exponential increases in computing power—culminating in the era of kilometer-resolution Earth system modeling—we still face significant challenges in accurately simulating certain meteorological phenomena. In this talk, we present our latest advancements in leveraging state-of-the-art supercomputers to further enhance weather and climate model capabilities. By tightly integrating high-performance computing (HPC)-based numerical models with data-driven AI methodologies, our approach aims to improve predictive accuracy for extreme events and address particularly challenging forecast windows, such as subseasonal and seasonal rainfall predictions. We believe that a robust, synergistic fusion of HPC and AI holds the promise of delivering groundbreaking breakthroughs in Earth system modeling. – Prof Fu Haohuan, Professor, Tsinghua University |
| 02:30pm – 02:45pm | Expanding HPC Capabilities with AI Innovations Research institutions around the world are building AI assets and cross-domain agents to implement complex scientific workflows. One of the largest challenges these institutions face is how to provide the performance users need at a sustainable cost and power level. SambaNova will discuss how a variety of research institutions are deploying production interference services based on SambaNova to address these challenges. – Jennifer Glore, VP of Customer Engineering, SambaNova Systems |
| 02:45pm – 03:30pm | Panel Discussion 1: “Advancing AI through Next-Generation HPC” In this panel, speakers from Session 1 will discuss future high-performance architectures, technologies and developments that will drive new capabilities in larges-scale AI. – Prof Satoshi Matsuoka – Marshall Choy – Jennifer Glore – Prof Fu Haohuan – Dr Ng Aik Beng – Moderated by Dr Rick Goh |
| 03:30pm – 04:00pm | Tea Break |
| 04:00pm – 04:05pm | Pitch 1: Multi-Agent AI Systems for Materials Innovation: Accelerating Discovery with Adaptive AI This presentation explores the integration of multi-agent AI systems in materials innovation. Dr. Liu will demonstrate how adaptive AI, combined with active learning and uncertainty quantification, can optimize material properties from sparse experimental datasets. The session will outline key challenges in materials R&D and showcase DeepVerse’s proprietary platform—featuring solutions such as Lab Assistant and Auto Lab—that delivers breakthrough efficiency gains. Attendees will gain insights into the future of computationally assisted materials research. – Dr Fredrik Liu, CEO and Co-founder, DeepVerse |
| 04:05pm – 04:10pm | Pitch 2: HPC and AI: A Synergistic Partnership for the Future of Computing This talk explores the synergistic relationship between High-Performance Computing (HPC) and Artificial Intelligence (AI), highlighting how this partnership is shaping the future of computing. HPC provides the computational power needed to train and deploy complex AI models, while AI algorithms optimize HPC systems, improving efficiency and performance. This synergy is driving breakthroughs across various fields, from scientific research and healthcare to finance and engineering. The talk will also discuss future trends, including AI-driven hardware architectures, edge computing, and quantum computing, and address the challenges and opportunities presented by this convergence. – Christopher Yeo, CEO and Founder, Sentient.io |
| 04:10pm – 04:15pm | Pitch 3: Meeting Users Where They Are, & Empowering Them To Go Where They Want, How They Want How Tenstorrent is using open standards, hardware, and software to enable the next generation of researchers – Felix Leclair, Field Application Engineer-HPC, Tenstorrent |
| 04:15pm – 04:30pm | Modeling and Simulation or AI? How will HPC look like in 10 years? AI is rapidly becoming the fastest emerging workload in HPC. However, traditional classical simulations, such as Molecular Dynamics, Engineering, and Weather and Climate modeling, continue to dominate the HPC landscape. This talk explores the future interplay between AI and classical simulation workloads. Will AI eventually supplant traditional methods in the medium term, or will we see the integration of AI-augmented simulations and simulation-augmented AI? We delve into the characteristics of such workloads across diverse fields, from Healthcare to Physical simulations, and examine scenarios where the synergy of AI and traditional simulations can be harnessed effectively, as well as instances where AI may completely take over. Join us as we outline the promising convergence of AI and classical simulations, paving the way for innovative advancements in HPC. – Prof Torsten Hoefler, Professor, ETH Zurich |
| 04:30pm – 04:45pm | Efficient Large-Scale Training and Inference on Wafer-Scale Clusters The Cerebras hardware and software stack is co-designed for efficient training and low-latency inference of large-scale models. Weight streaming execution allows distributed training in a strictly data-parallel form for models and clusters of arbitrary sizes, avoiding complex and time-consuming hybrid distribution techniques. A large pool of on-chip memory enables ultra-low-latency autoregressive inference. Leveraging our experience in training large language models (LLMs) and multi-modal models, we will share insights into optimizing model architectures and training strategies for compute-efficient training. Additionally, we will explore hardware-optimized LLM mapping to wafer-scale clusters for low-latency autoregressive inference. – Dr Natalia Vassilieva, VP and Field CTO, Cerebras |
| 04:45pm – 05:00pm | Hybrid Quantum-Classical Computing for Next-Gen AI Quantum computing is poised to revolutionize supercomputing and its applications, particularly through enabling the development of next-gen AI with quantum machine learning (QML). Novel QML models can be developed by leveraging near-term GPU-QPU hybrid systems through GPU emulation and hybrid variational quantum approaches utilizing both classical and quantum hardware resources. Recent advancements demonstrate the potential of QML to achieve precision and generalizability on par with, or surpassing, classical models in processing real-world data and tasks. This talk will explore the integration of quantum computing with GPUs, highlighting the transformative potential of hybrid quantum-classical systems in developing next-gen AI. – Dr Jie Luo (Roger), Co-Founder and CEO, Anyon Technologies |
| 05:00pm – 05:45pm | Panel Discussion 2: “Empowering Industry Technologies through HPC and AI” In this panel, speakers from Session 2 will discuss the ongoing developments and challenges in bridging HPC-AI and essential use cases across science and industry. We will explore the role of HPC-AI as a catalyst in areas including climate science, modelling and simulation, quantum computing. – Prof Torsten Hoefler – Dr Natalia Vassilieva – Dr Jie Luo (Roger) – Dr Liu Yong – Moderated by Dr Yang Liwei |
| 05:45pm | Closing |
