Siva Rajamanickam is a Distinguished Member of Technical Staff in the Center for Computing Research at Sandia National Laboratories and national co-lead for infrastructure on the Genesis Mission Science and Security Platform. In that role he co-leads the software and hardware infrastructure pillar across 17 DOE national laboratories, building DOE’s federated AI platform spanning on-prem compute, multi-cloud resources, and experimental facilities — including the closed-loop agentic systems behind the Aires Tide demonstration.
Siva leads a distributed technical organization across national laboratories spanning model development, infrastructure, and co-design, and serves as Tri-Labs lead for Chandler, NNSA’s first federated foundation model, and for the co-design program with Cerebras Systems. He co-founded the Kokkos Kernels and Kokkos ecosystem projects underpinning the U.S. exascale software stack and has led Trilinos linear solvers and its strategic leadership team. Siva’s recognitions include the Secretary of Energy’s Honor Award, a Gordon Bell finalist selection, and two R&D 100 awards.
Presentation Title:
Genesis Mission Platform and National Security Applications
Presentation Abstract:
The Genesis Mission Science and Security Platform connects models, data, and computing infrastructure into a single federated environment spanning 17 national laboratories, accelerating discovery across the Department of Energy’s science and security missions. This talk describes the platform’s infrastructure design and the closed-loop agentic workflows it enables, in which reasoning engines and frontier models generate hypotheses, drive simulations and experiments, and learn from the results.
The talk then turns to what is distinct about building such a platform for the national security mission. Sensitive and classified data impose requirements on data isolation, and agents operating on national security problems raise issues that general-purpose AI infrastructure does not address. Aires Tide, which produced a flight test vehicle in under five months, illustrates the end-to-end capability of the platform and its applicability to national security.