SPADE 2025: 1st International Workshop on Scheduling & Parallelism in AI for Distributed Edges

As cloud, edge, and mobile computing continue to converge into a unified computing ecosystem, there is a growing opportunity for edge platforms to support computationally intensive AI models as services similar to the cloud AI services. However, constraints on scheduling, distribution, and resource managements making large scale adoption slow. Considering these challenges textbf{Scheduling and Parallelism for AI in Distributed Edge } (SPADE) aims to collect novel contributions on the architecture, algorithmic, and practical challenges of deploying computationally intensive applications (Deep learning, Generative AI, LLM, etc) across distributed edge computing. While edge AI offers low-latency and privacy-preserving inference for IoT, the inability to efficiently parallelize tasks often limits system performance. This workshop targets the combination of computationally intensive models handling parallelism limits, AI computation and scheduling optimization with resource-constrained, and heterogeneous edge networks. SPADE will bring together researchers and practitioners working on model partitioning, decentralized scheduling, execution frameworks, hybrid AI pipelines, and benchmarking testbeds. In addition, it will address bottlenecks in distributed inference and highlight new strategies for intelligent scheduling and task coordination. The contributions on model slicing, distributed inference, workload orchestration, task-to-node mapping, and scalable deployment of AI for IoT systems through presentations, discussion sessions, and short papers is welcome.

The thematic coverage of the event includes, but is not restricted to, the following topics: task distribution and scheduling, edge ai and tinyml, hybrid edge cloud execution frameworks and testbeds for edge based ai.