Glia: A Human-Inspired AI for Automated Systems Design and Optimization

2025/10/31 by Pouya Hamadanian, Pantea Karimi, Hamadanian, Pouya +13 · 1 voice · 5 citations
Computer Science · #Artificial Intelligence (cs.AI) #Computation and Language (cs.CL) #Distributed #FOS: Computer and information sciences #Parallel #and Cluster Computing (cs.DC) #cs.AI #cs.CL #cs.DC

paper · pdf · doi:10.48550/arxiv.2510.27176

arxiv published 2025/10/31 · arxiv updated 2026/04/03

Abstract

Can 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. 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 combining reasoning LLMs with structured experimentation, an AI can produce creative and understandable designs for complex systems problems.

Citations

• SR-Scientist: Scientific Equation Discovery With Agentic AI
• Man-Made Heuristics Are Dead. Long Live Code Generators!
• Scientific Algorithm Discovery by Augmenting AlphaEvolve with Deep Research
• MetaMuse: Algorithm Generation via Creative Ideation
• Astra: A Multi-Agent System for GPU Kernel Performance Optimization
• An AI system to help scientists write expert-level empirical software
• Congestion Control System Optimization with Large Language Models
• \(X\)-evolve: Solution space evolution powered by large language models
• AlphaGo Moment for Model Architecture Discovery
• AlgoTune: Can Language Models Speed Up General-Purpose Numerical Programs?
• GPU Kernel Scientist: An LLM-Driven Framework for Iterative Kernel Optimization
• AlphaEvolve: A coding agent for scientific and algorithmic discovery
• CALM: Co-evolution of Algorithms and Language Model for Automatic Heuristic Design
• Fitness Landscape of Large Language Model-Assisted Automated Algorithm Search
• Current and Future Use of Large Language Models for Knowledge Work
• Monte Carlo Tree Search for Comprehensive Exploration in LLM-Based Automatic Heuristic Design
• LLM4AD: A Platform for Algorithm Design with Large Language Model
• Improving Parallel Program Performance with LLM Optimizers via Agent-System Interfaces
• Towards Safer Heuristics With XPlain
• Multi-objective Evolution of Heuristic Using Large Language Model
• Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters
• A Queueing Theoretic Perspective on Low-Latency LLM Inference with Variable Token Length
• Vidur: A Large-Scale Simulation Framework For LLM Inference
• Designing Network Algorithms via Large Language Models
• Taming Throughput-Latency Tradeoff in LLM Inference with Sarathi-Serve
• Evolution of Heuristics: Towards Efficient Automatic Algorithm Design Using Large Language Model
• Mathematical discoveries from program search with large language models
• Splitwise: Efficient generative LLM inference using phase splitting
• Efficient Memory Management for Large Language Model Serving with PagedAttention
• Lost in the Middle: How Language Models Use Long Contexts
• Reparo: Loss-Resilient Generative Codec for Video Conferencing
• Training Verifiers to Solve Math Word Problems
• Adaptive Neural Signal Detection for Massive MIMO
• Real-world Video Adaptation with Reinforcement Learning
• Scaling Distributed Machine Learning with In-Network Aggregation
• Internet Congestion Control via Deep Reinforcement Learning
• Learning Scheduling Algorithms for Data Processing Clusters
• DeepConfig: Automating Data Center Network Topologies Management with Machine Learning
• RLWS: A Reinforcement Learning based GPU Warp Scheduler
• Reinforced Generation of Combinatorial Structures: Hardness of Approximation
• Automatically discovering heuristics in a complex SAT solver with large language models

Cited by

• A Vocabulary for Multi-Agent Automated Research Systems
• Automated Design Optimization via Strategic Search with Large Language Models
• Improving Coherence and Persistence in Agentic AI for System Optimization
• TuxBot: Semantic-Aware Online OS Tuning with Large Language Models
• Vulcan: Instance-specialized, Verifiable Systems Heuristics Through LLM-driven Search

Discussions

• Research introduces "Glia," a framework aimed at automated systems design and optimization. It looks like Glia is an AI approach intended to mimic human design intuition for complex systems. arxiv.org [bsky, 2 points, 1 comments]

Related

• David vs. Goliath: Can Small Models Win Big with Agentic AI in Hardware Design?
• Barbarians at the Gate: How AI is Upending Systems Research
• AI for Distributed Systems Design: Scalable Cloud Optimization Through Repeated LLMs Sampling And Simulators
• Can Reasoning Models Reason about Hardware? An Agentic HLS Perspective
• BO-Muse: A human expert and AI teaming framework for accelerated experimental design