Maroon Ayoub

Maroon Ayoub is a Staff Research Scientist at IBM Research, focused on AI systems, distributed inference infrastructure, and Kubernetes-native model serving. His work bridges large-scale systems design with production AI, leading efforts in KV-Cache management, prefix-aware routing, and distributed scheduling for high-performance LLM workloads.

He leads the Disaggregated-KV Special Interest Group (SIG) and serves as a core contributor to the open-source llm-d project, driving innovations in cache-aware routing, intelligent KV-cache management & optimizations, and adaptive scheduling for vLLM-based clusters. Earlier in his career, he played a leading role in KubeStellar, an IBM initiative on multi-cluster scheduling and policy-based workload placement.

Maroon also supervises industry-linked student projects at the Technion – Israel Institute of Technology, mentoring teams at the intersection of research and real-world systems development.