Rahul K

At Cedana, I build managed Kubernetes infrastructure for Caltech, supporting 6,000+ CPU and 2,000+ GPU workloads on EKS with cedana-cli.

I implemented spot-interruption checkpoint and restore workflows through a custom Cedana operator, improving performance by up to 80% and reducing monthly costs by more than $5,000. I also built a Rust policy engine for custom Kubernetes checkpoint and restore actions and durable heartbeat checkpoint execution through Rust microservices.

I’ve developed HPC infrastructure on EKS with Kueue and Slurm, orchestrated PyTorch and molecular dynamics/co-folding workloads, and enabled live migration for CUDA applications. I also created an H100 testing framework for Kubernetes checkpoint-restore operations across OCI runtimes.

Previously at CloudSEK, I supported releases and infrastructure for applications scanning 3,000,000 websites, maintained ArgoCD GitOps and Helm workflows, and built EKS and GKE infrastructure. I enjoy working across Rust, Kubernetes, cloud platforms, observability, and systems reliability.