AIPerf 2027

Modern software systems are becoming increasingly complex, while Artificial Intelligence (AI) is rapidly transforming how software is developed, operated, and optimized. At the same time, AI systems themselves have become large-scale, resource-intensive, and performance-critical software systems.

AIPerf: Workshop on AI for Software Performance and Performance Engineering for AI Systems focuses on the intersection of AI and performance engineering from two complementary perspectives: using AI to improve the analysis, prediction, diagnosis, testing, tuning, and optimization of software performance, and applying performance engineering techniques to AI systems to improve their efficiency, reliability, scalability, cost, and sustainability.

AIPerf aims to bring together researchers and practitioners from software engineering, performance engineering, AI systems, AIOps, MLOps/LLMOps, cloud computing, and sustainable computing to share emerging results, exchange practical experiences, discuss open challenges, and identify future research directions at this rapidly growing intersection.

Call for Papers

AIPerf invites research papers, short papers, position papers, experience reports, tool and dataset papers, and non-archival lightning talk proposals on the intersection of artificial intelligence, software engineering, and performance engineering.

Modern software systems are increasingly complex and are expected to deliver strong performance, reliability, cost efficiency, and energy efficiency. At the same time, software teams have growing access to rich telemetry from tests, profiling, tracing, logs, build histories, configuration repositories, and production operations. These data create new opportunities for AI techniques to support performance engineering by modeling performance behavior, predicting regressions, localizing bottlenecks, tuning configurations, optimizing code and architectures, generating tests, reducing cloud cost, and improving energy efficiency. Conversely, AI systems themselves are performance-critical software systems: training, inference, serving, retrieval, agentic workflows, and AI operations require principled engineering for latency, throughput, utilization, reliability, cost, and sustainability.

Topics of Interests:

• AI-assisted performance testing, benchmarking, regression detection, and load generation.
• Learned performance models, bottleneck localization, root-cause analysis, and observability.
• Automated tuning, resource management, architecture optimization, and code optimization.
• Energy-aware and cost-aware software optimization.
• LLM agents and human-AI workflows for performance engineering.
• Performance engineering for AI training, inference, serving, RAG, and agentic systems.
• MLOps and LLMOps practices for performance, reliability, and cost control.
• Benchmarks, datasets, artifacts, negative results, and industrial experience reports.

Submission types

• Research papers may be up to 8 pages excluding references.
• Position papers may be up to 5 pages excluding references.

All submissions must follow the final ICSE 2027 workshop formatting and proceedings instructions.

Review and publication

Accepted abstracts will be published in the IEEE workshop proceedings, and every submission will receive at least three reviews. The official publication date of the workshop proceedings is the date the proceedings are made available by IEEE. This date may be up to two weeks prior to the first day of ICSE 2027. The official publication date affects the deadline for any patent filings related to published work.

At least one author of each accepted paper must register and present the work.