SI² — SuperIntelligence for Science

Trustworthy analytics for HPC operations

Introducing two research directions in multi-agent operational analytics and telemetry-informed decisions.

Research perspective · SI² — SuperIntelligence for Science

Beyond retrieving the right records

Recall Is Not Honesty examines multi-agent architectures for trustworthy HPC operational analytics. The work has been accepted at SC ’26, as reported by the research team.

Toward defensible decisions

EPIC++ is ongoing research into decentralized multi-agent analytics for job submission decisions on HPC telemetry. It explores how operational evidence can support defensible choices.

Research across institutions

The supplied author teams connect Oak Ridge National Laboratory, Predii, Lawrence Berkeley National Laboratory, spinTwo, Rensselaer Polytechnic Institute, and AMD.

Explore related research →