APWA: A Distributed Architecture for Parallelizable Agentic Workflows

2026-05-14

AI Summary

Existing multi-agent systems struggle to efficiently leverage the parallel capabilities of large language models in large-scale, complex tasks due to substantial reasoning coordination overhead and computational scalability bottlenecks. This work proposes a distributed multi-agent architecture that dynamically decomposes tasks into independent subproblems, enabling parallel execution without inter-node communication and supporting heterogeneous data and diverse parallelization strategies. The approach introduces, for the first time, a communication-free parallelization mechanism tailored for highly parallelizable agent workflows. It significantly enhances system throughput and scalability on complex query tasks and remains effective even in large-scale scenarios where prior methods completely fail.

Abstract

Autonomous multi-agent systems based on large language models (LLMs) have demonstrated remarkable abilities in independently solving complex tasks in a wide breadth of application domains. However, these systems hit critical reasoning, coordination, and computational scaling bottlenecks as the size and complexity of their tasks grow. These limitations hinder multi-agent systems from achieving high-throughput processing for highly parallelizable tasks, despite the availability of parallel computing and reasoning primitives in the underlying LLMs. We introduce the Agent-Parallel Workload Architecture (APWA), a distributed multi-agent system architecture designed for the efficient processing of heavily parallelizable agentic workloads. APWA facilitates parallel execution by decomposing workflows into non-interfering subproblems that can be processed using independent resources without cross-communication. It supports heterogeneous data and parallel processing patterns, and it accommodates tasks from a wide breadth of domains. In our evaluation, we demonstrate that APWA can dynamically decompose complex queries into parallelizable workflows and scales on larger tasks in settings where prior systems fail completely.

Problem

• multi-agent systems
• parallelizable workflows
• computational scaling
• reasoning bottlenecks
• coordination challenges

Innovation

• distributed architecture
• parallelizable workflows
• multi-agent systems
• LLM-based agents
• task decomposition