1st Workshop on Systems for Agentic AI

As generative and agentic AI systems grow more powerful and widely adopted, the need for efficient, scalable infrastructure to support their deployment has become more pressing than ever. This workshop aims to bring together leading researchers and practitioners from both academia and industry to share insights, challenges, and breakthroughs in building high-performance systems for generative and agentic AI. By fostering collaboration between system designers, infrastructure engineers, and AI researchers, we seek to advance the state of the art in deploying AI systems that are cost-effective, responsive, and capable of interacting with complex real-world environments.

Call for Presentations

We invite authors to submit an abstract outlining their proposed presentation. We welcome new challenges, system designs, optimization techniques, or lessons learned from production deployments of generative or agentic AI.

Topics of Interests

• Foundations of Generative and Agentic AI Inference
• Architectural Patterns and Design Considerations for AI Systems
• Cost- and Latency-Optimized Inference Techniques
• System Architectures for Serving AI Agents
• Case Studies from Real-World Deployments
• Bottleneck Identification and Performance Tuning in Model Serving
• Model Compression (Quantization, Pruning, Distillation)
• Caching, Batching, and Speculative Decoding for Throughput Optimization
• Hardware-Aware and System-Level Optimizations (e.g., FlashAttention, Scheduler Tuning)
• Efficient Inference on Specialized Hardware (GPUs, TPUs, Neuron, Custom ASICs)
• Deployment Strategies for Large-Scale Agentic Systems
• Evaluation Methodologies for Inference Optimization
• Scaling Laws and Trends in Inference Workloads

Important Dates

• Submission deadline: August 7, 2025 → August 14, 2025
• Acceptance notification: August 23, 2025
• Camera-ready due: October 3, 2025
• Workshop date: October 13, 2025

Submission Guidelines

Submissions should be no more than two pages in length, excluding references, and must follow the standard ACM two-column conference format with a 10-point font. Authors are encouraged to consult the official ACM website for the most up-to-date template: https://www.acm.org/publications/proceedings-template.

Submissions will be reviewed in a single-blind process. Presentations will be selected based on their potential to spark insightful technical discussions at the workshop and to influence future systems research. Evaluation criteria include originality, technical merit, clarity, relevance, industry impact, and the depth of insights or lessons shared. Please note that submissions will not be published in the ACM Digital Library.

Submissions can be made online at https://saa25.hotcrp.com.

Accepted Papers

Authors of accepted papers will be expected to supply electronic versions of their papers. Camera-ready papers may be up to 4 pages in length, excluding references. To facilitate broad technical discussion, all accepted papers will be made available online in advance of the workshop.

Program Schedule

To be announced

Organizers

• Byung-Gon Chun, FriendliAI and Seoul National University
• Jae Wook Lee, Seoul National University
• Gyeong-In Yu, FriendliAI

Contacts

Any organizational questions or submission inquiries can be emailed to saa25@googlegroups.com.

Sponsorship Opportunities

We welcome sponsors interested in promoting their AI infrastructure tools, model inference services, AI agent services, or cloud platforms. Please contact the organizers at saa25@googlegroups.com.

Sponsors