Work

Towards Resource-Efficient Compound AI Systems

Year: 2025

Type: conference-paper

Abstract: Compound AI Systems, integrating multiple interacting components like models, retrievers, and external tools, have emerged as essential for addressing complex AI tasks. However, current implementations suffer from inefficient resource utilization due to tight coupling between application logic and execution details, a disconnect between orchestration and resource management layers, and the perceived exclusiveness between efficiency and quality.

Source: Workshop on Hot Topics in Operating Systems

DOI: 10.1145/3713082.3730377

Authors: Gohar Irfan Chaudhry, Esha Choukse, Íñigo Goiri, Rodrigo Fonseca, Adam Belay, Ricardo Bianchini

Institutions: MIT Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology, Microsoft (United States)

Language: English

Topic: Software System Performance and Reliability

Subfield: Computer Networks and Communications

Field: Computer Science

Domain: Physical Sciences

Open Access status: hybrid

Funders: Semiconductor Research Corporation, Defense Advanced Research Projects Agency

Award: G1452086972