Preprint
Jul 2026
A Workflow-Aware Serving Layer for Agentic Applications
Dyserve compiles each workflow's per-node model and verifier choices in one integer linear program (ILP) over a heterogeneous backend pool, priced by skill-conditioned offline profiles that transfer across workflows.
Jiayi Qian, Zishen Wan, Hanchen Yang et al.
· 0 citations