Dr.Seelam Seetharami
Dr. Seetharami R. Seelam is an IBM Distinguished Engineer and Chief Architect of Classical System Co-Design for Quantum-Centric Supercomputing (QCSC). His work spans high-performance computing, AI, hybrid cloud, and quantum computing, with a focus on building integrated quantum-classical systems and scaling the classical components needed for future quantum workloads. He leads efforts in GPU-based acceleration, resource scheduling, and reference architectures that bring together quantum, HPC, and AI systems, while collaborating with national laboratories and industry partners.
Previously, he led global research teams at IBM on HPC, AI, cloud infrastructure, networking, storage, accelerators, and distributed resource management. Dr. Seelam has also contributed to academia through teaching at New York University and Columbia University. He has filed more than 30 patents, published over 35 papers, received multiple paper awards, and is a frequent speaker at major academic and industry conferences.
How will quantum computers and HPC and AI systems work together? Must they be co-located and tightly coupled, or can geographically distributed systems still enable workloads to leverage the unique capabilities of each platform? What scheduling and resource management challenges emerge when quantum systems become part of HPC datacenters and are integrated into HPC workflows? These are questions we at IBM have been addressing in a recent paper proposing a reference architecture for a Quantum-Centric Supercomputer.
In this talk, we present a nuanced view — grounded in workload requirements and technology maturity — of how quantum and HPC systems will work together across near- and long-term timescales. We propose a co-designed strategy spanning quantum and classical HPC infrastructure, middleware, and application layers to accelerate quantum adoption for critical computational problems. We frame this Quantum-Centric Supercomputing (QCSC) evolution in three phases: (1) quantum systems as specialized compute offload engines within existing HPC complexes; (2) heterogeneous quantum-classical systems coupled through advanced middleware, enabling seamless hybrid algorithm execution; and (3) fully co-designed quantum- HPC systems purpose-built for hybrid computational workflows. We conclude with a reference architecture, roadmap and an update on realizing each of these phases.
