Cedana tackles the critical scarcity of AI and HPC resources by introducing automated GPU checkpointing infrastructure. This allows seamless migration of workloads across instances without data loss, directly addressing the high costs and time losses from cluster failures. The system operates at the kernel/OS level, requiring no code changes, and integrates with Kubernetes, SLURM, and NVIDIA Dynamo. By maximizing utilization and reliability, Cedana accelerates research output and revenue for clients.

The company’s technical foundation is built on a decade of expertise in computation efficiency. Cedana’s team has published research at NeurIPS and CVPR, including early work on distributed training convergence. Their experience spans Shopify’s warehouse automation and enterprise HPC deployments. The solution’s compatibility with major platforms like Kubernetes and SLURM ensures broad applicability, solving pain points in infrastructure complexity.

As a Forward Deployed Engineer at Cedana, you’ll own end-to-end client deployments, from SLURM at universities to Kubernetes in enterprise environments. The role involves debugging production issues, optimizing migration policies, and scaling installation playbooks. This hands-on approach ensures Cedana’s platform adapts to diverse infrastructures while minimizing client overhead. The company’s pause/migrate/resume compute paradigm represents a systems-level shift in resource allocation for HPC and AI, working directly with hardware at the Linux kernel layer.