Built cloud-native and AI infrastructure across gateways, service mesh, and Kubernetes.
Open AI Infrastructure · Mixture-of-Models
Xunzhuo Liu
Building the system of many models.
Right workload. Right model. Right hardware.
The future is not one model.
It is a system of models.
Applications choose the objective; semantic routing chooses the execution path across models, tools, memory, runtimes, and hardware. My focus is preference-aligned Mixture-of-Models, inspectable decisions, and Workload–Router–Pool co-design.
Read The Semantic Routing Moment
Open Source
From cloud infrastructure to system intelligence.
The same systems thread—from cloud-native foundations to open model infrastructure.
Senior Member of Technical Staff
Open Source & AI
Signature Project
vLLM Semantic Router
Co-founder & Core Maintainer
The open semantic decision layer for model routing, bounded model collaboration, and Mixture-of-Models systems.
Selected open-source roles · past & present
A path from service mesh and Kubernetes gateways to open AI inference infrastructure.
Envoy Gateway
MaintainerKubernetes and standalone application gateway built on Envoy Proxy.
Envoy AI Gateway
MaintainerOpen AI traffic layer for provider access, routing, and inference controls.
vLLM AIBrix
MaintainerCloud-native building blocks for scalable, cost-efficient LLM inference.
Istio
MaintainerService mesh control plane for connecting, securing, and observing services.
Kiali
MaintainerObservability and management console for Istio service meshes.
Aeraki Mesh
MaintainerLayer-7 protocol management for service-mesh workloads.
Higress
CommitterCloud-native AI gateway and API gateway.
Kubernetes Gateway API
Member & ReviewerPortable, role-oriented Kubernetes APIs for traffic and gateway infrastructure.
Kubernetes WG AI Gateway
Founding Member and ChairCommunity work defining portable AI Gateway patterns for Kubernetes.
Merbridge
MaintainereBPF acceleration for service-mesh data paths.