Thrust 4: Driving Applications

decorative image with a bar graph and arrow trending up viewed through a magnifying glass

This thrust benchmarks sustainability across high-impact application areas — particularly artificial intelligence — while studying the trade-offs between embodied and operational efficiency in domain-specific architectures. By profiling real workloads from AI training and inference, this thrust generates the use cases and design constraints that ground the center’s research.

Key Research Questions

  • How do model architecture choices shift the balance between embodied and operational costs?
  • When do domain-specific accelerators’ embodied costs outweigh their operational savings?
RolePIInstitution
LeadEmma StrubellCMU
CollaboratorDavid BrooksHarvard
CollaboratorUdit GuptaCornell
CollaboratorBenjamin LeePenn