KAIROS: Stateful, Context-Aware Power-Efficient Agentic Inference Serving

TL;DR

KAIROS is a novel, context-aware system designed to optimize power efficiency in agentic AI inference serving by dynamically managing GPU resources based on long-lived context and agent progress.

Contribution

It introduces a new power management approach that considers agent context and request evolution, outperforming traditional single-turn optimization techniques.

Findings

Achieves 27% average power reduction while maintaining performance.

Effectively manages GPU frequency and request placement based on agent context.

Reduces power consumption across diverse agentic AI tasks.

Abstract

Power has become a central bottleneck for AI inference. This problem is becoming more urgent as agentic AI emerges as a major workload class, yet prior power-management techniques focus almost entirely on single-turn LLM serving. Our analysis shows that agentic serving behaves fundamentally differently: each request carries long-lived context that evolves across tool-interleaved turns, and lowering GPU frequency can push the system into a thrashing regime where memory pressure sharply worsens both performance and power efficiency. These observations show that power optimization for agentic serving requires rethinking. We present KAIROS, a context-aware power optimization system for agentic AI serving. KAIROS uses agent context as a first-class control signal to jointly manage GPU frequency, per-instance concurrency, and multi-instance request placement. This enables KAIROS to save power when memory headroom exists while avoiding thrashing and preserving performance targets. At a high level, KAIROS tracks requests at agent granularity, adapts local control to context growth and agent progress, and routes agents across instances to jointly improve power efficiency and memory stability. Evaluated across diverse software and data engineering agentic tasks, KAIROS achieves an average of 27% (up to 39.8%) power reduction while meeting the performance targets.