About the Author#
With more than 25 years in enterprise and cloud architecture, Raj has worked with Kubernetes since 2015 and today focuses on GPU and HPC infrastructure for AI: accelerators, RDMA fabrics, collective communication, and the Kubernetes control plane that runs it all.
His work has spanned Fortune 500 organizations in finance, aerospace, government and technology.
Focus areas#
- GPU clusters on Kubernetes: GPU Operator, device plugins and DRA, scheduling with Kueue and Volcano
- High-performance networking: InfiniBand, RoCE v2, HPE Slingshot, RDMA into pods with Multus
- Collective communication: NCCL and libfabric, on which he has written technical papers
- Multi-tenant platform design: isolation, quotas and GPU sharing
- Custom layered OS design for GPU Kubernetes nodes
Experience and education#
- Cloud: 15+ years architecting and delivering enterprise applications and platforms on public, private and hybrid cloud
- Kubernetes: 10+ years designing and implementing microservices-based platforms on Kubernetes
- Education: M.S. Computer Science; B.E. Mechanical Engineering
Also by Raj#
K8AI: the AI stack, layer by layer ↗: an open knowledge base for building GPU platforms on Kubernetes, from the AI workload through parallelism and NCCL down to RDMA fabrics, NVLink and silicon.