About the Workshop
The rapid evolution of agentic and multimodal AI systems is redefining how applications, algorithms, and infrastructure co-evolve. Yet, current computing stacks—spanning programming languages, operating systems, compilers, networks, and architectures—remain largely siloed, creating inefficiencies in scalability, latency, and cost.
The Co-Design for Agentic and Multimodal AI (CoDAIM 2026) workshop aims to bring together researchers and practitioners from programming languages, compilers, systems, and computer architecture domains to explore principled co-design approaches that unify these layers. Topics include systems for large-scale AI inference and agentic workloads, architectural patterns for efficient deployment, model compression and scheduling strategies, and novel use of hardware/software co-optimization for multimodal reasoning and interaction.
The workshop will feature invited keynotes and contributed talks to identify emerging challenges and opportunities at the intersection of AI systems and architecture. We encourage submissions presenting early-stage ideas, position papers, or empirical studies that highlight bottlenecks, propose cross-stack solutions, or examine scaling trends in AI inference.
Accepted papers will be made openly available, and a post-workshop summary paper will capture key insights, discussions, and directions for the broader ASPLOS community. Workshop artifacts will also serve as open educational and research resources to foster collaboration across disciplines.
Important Dates
Workshop Highlights
- Keynotes: Leading researchers from academia and industry presenting frontier perspectives
- Contributed Talks: Selected papers with 12-min presentation + 3-min Q&A
- Post-Workshop Summary: Key insights will be published on arXiv or ACM Digital Library