Glia: A Human-Inspired AI for Automated Systems Design and Optimization
🤖 AI Summary
Can AI autonomously design computer systems comparable to human experts? This work proposes a human-like multi-agent architecture that integrates large language models’ abstract reasoning with empirical feedback to enable interpretable, iterative automated system design. Methodologically, it decouples reasoning, experimentation, and analysis modules, establishing a structured evaluation framework and closed-loop feedback mechanism—overcoming the limitations of traditional black-box optimization. Applied to distributed GPU clusters, the system autonomously generates high-performance request routing, scheduling, and elastic scaling algorithms; achieves design quality on par with human experts; reduces design time by an order of magnitude; and discovers, for the first time, novel patterns in workload evolution. This is the first demonstration of AI-driven, end-to-end system-level design that simultaneously exhibits creativity, transparency, and engineering practicality.
Application Category
📝 Abstract
Can an AI autonomously design mechanisms for computer systems on par with the creativity and reasoning of human experts? We present Glia, an AI architecture for networked systems design that uses large language models (LLMs) in a human-inspired, multi-agent workflow. Each agent specializes in reasoning, experimentation, and analysis, collaborating through an evaluation framework that grounds abstract reasoning in empirical feedback. Unlike prior ML-for-systems methods that optimize black-box policies, Glia generates interpretable designs and exposes its reasoning process. When applied to a distributed GPU cluster for LLM inference, it produces new algorithms for request routing, scheduling, and auto-scaling that perform at human-expert levels in significantly less time, while yielding novel insights into workload behavior. Our results suggest that by combining reasoning LLMs with structured experimentation, an AI can produce creative and understandable designs for complex systems problems.
Problem
Research questions and friction points this paper is trying to address.
AI autonomously designs computer systems mechanisms
Generates interpretable designs with exposed reasoning process
Produces human-expert level algorithms for distributed systems
Innovation
Methods, ideas, or system contributions that make the work stand out.
Multi-agent workflow with specialized reasoning roles
Combining LLM reasoning with empirical feedback framework
Generating interpretable designs instead of black-box policies
🔎 Similar Papers
No similar papers found.
