What problem does it solve? Platform teams running GPU research workloads on Kubernetes need every add-on and tenant environment declared in Git and continuously reconciled, instead of hand-applied kubectl changes that drift, race during bootstrap, and bypass review. ## Core Features & Use Cases - App-of-apps bootstrap: Stand up a root Argo CD Application that fans out to platform add-ons (GPU operator, DCGM monitoring, KEDA, JupyterHub, model serving) from a single repo. - ApplicationSet tenant vending: Generate one Application per folder under tenants/ so onboarding a researcher is merging a PR, with prune-based deprovisioning. - Sync-wave ordering and secrets strategy: Order CRDs, GPU operator, monitoring, and workloads with sync waves, and choose between Sealed Secrets, SOPS, and External Secrets Operator. - Use Case: On a fresh Lambda-managed Kubernetes cluster, install Argo CD, apply the root app, and have the full GPU platform plus per-researcher namespaces reconcile from Git with drift self-healing. ## Quick Start Ask the agent to design an Argo CD app-of-apps bootstrap repo with sync waves and an ApplicationSet that vends one namespace per tenant folder for your GPU cluster.