AIRS 2026

About the Workshop

Agentic AI is moving from model-level reasoning benchmarks and single-turn LLM applications into deployed systems that coordinate tools, models, data, services, and people under real operational constraints. These systems increasingly combine LLM inference, retrieval, code execution, external APIs, databases, tool sandboxes, multi-agent coordination, and human-in-the-loop workflows.

AIRS treats agentic AI as a first-class distributed systems workload: dynamic, stateful, tool-using, long-running, and deployed under real resource, reliability, and security constraints. The workshop aims to bring together researchers and practitioners from parallel and distributed computing, ML systems, LLM serving, multi-agent systems, security, and trustworthy AI.

Non-Archival Workshop

AIRS 2026 is planned as a non-archival workshop. Accepted submissions will be presented at the workshop for discussion and feedback, but will not be included in formal proceedings. This format is intended to encourage early-stage work, work-in-progress results, vision papers, and open discussion.

Call for Papers

AIRS 2026 welcomes regular research papers, short papers, work-in-progress papers, and vision papers on systems, infrastructure, algorithms, and deployment for agentic AI in real-world settings. We especially encourage submissions that treat agentic AI as a systems problem, connecting model inference, tools, memory, state, orchestration, scheduling, and deployment constraints.

Topics include distributed runtime systems for agents, LLM serving and inference backends, multi-agent workflow execution, secure tool invocation, systems-oriented planning and scheduling, benchmarking, and trustworthy deployment. Accepted submissions will be presented for discussion and feedback in a non-archival workshop format and will not be included in formal proceedings.

Read the full Call for Papers Submit via OpenReview

Workshop Focus

The workshop is organized around three connected dimensions:

• Infrastructure and systems: scalable runtime support, orchestration, state management, secure tool execution, and cloud/edge deployment.
• Algorithms for agents: planning, scheduling, coordination, adaptation, memory management, and reliability-aware runtime decisions.
• Real-world deployment: evaluation and operational experience under performance, cost, privacy, safety, and reliability constraints.

Important Dates

All listed times use (GMT +8:00) Beijing, Perth, Singapore, Hong Kong unless otherwise specified.

Topics of Interest

AIRS welcomes work on systems, infrastructure, algorithms, and deployment for agentic AI in real-world settings. Topics include, but are not limited to:

• Infrastructure, orchestration, and distributed runtime systems for agentic AI
• LLM serving, inference backends, memory/state management, and accelerator-aware execution for agent workloads
• Scheduling, placement, resource management, and workflow execution for multi-agent systems
• Communication protocols, coordination mechanisms, tool invocation, and secure execution environments
• Systems-oriented algorithms for planning, scheduling, adaptation, and reliability-aware runtime decisions in agentic systems
• Benchmarking, workload characterization, and performance-cost-quality evaluation for deployed agents
• Trustworthy, private, and secure deployment of agentic AI systems in real-world environments

Workshop Format

AIRS is planned as a compact half-day program with invited talks, peer-reviewed contributed presentations, short talks or lightning talks when appropriate, and a panel discussion on the systems roadmap for deployable agentic AI.

• Invited talks on agent systems infrastructure, reliability, security, and deployment.
• Peer-reviewed presentations for accepted regular, short, work-in-progress, and vision submissions.
• A moderated panel discussion on open systems problems and evaluation gaps.

Organizers

• Qinbin Li, Huazhong University of Science and Technology
• Jialin Li, National University of Singapore
• Wei Dong, Nanyang Technological University
• Chaoyi Ruan, National University of Singapore
• Jing Gu, Huazhong University of Science and Technology