Builders series
How-to guides and tech showcases for building on Klique.
Fine-tune on the HPC cluster, serve on Kubernetes
Meta-scheduling in practice: one Python task fine-tunes an open-weight model on a Slurm cluster and on Kubernetes, the model registry hands it over by ID, and vLLM serves it on Kubernetes, all from one Klique control plane.
Build a physical AI data factory with NVIDIA Cosmos
NVIDIA Cosmos AnomalyGen generates the rare defects a production line seldom yields, a quality gate keeps the ones that measure up, and a coding agent drives the whole loop on Klique from a goal written in plain English.
