NVIDIA Kubernetes Engine (NVKE)

Kyverno is a Kubernetes-native policy engine for platform engineering teams that manages security, compliance, automation, and governance through policy as code. It uses Kubernetes admission controls and background scans to validate, mutate, generate, and clean up resources, and can verify container image signatures for supply-chain security. Kyverno works with tools including kubectl, Kustomize, and Git; its policies can enforce standards such as Pod Security Standards, security contexts, naming conventions, resource limits, image sources and signatures, and required workload probes. It complements Kubernetes RBAC and native ValidatingAdmissionPolicies and MutatingAdmissionPolicies rather than replacing them, and only enforces requirements explicitly defined in maintained policies.

KAI Scheduler is a scheduler from NVIDIA for Kubernetes-based AI workloads. It brings gang scheduling, which schedules a workload's required tasks together, and topology awareness to the Kubernetes ecosystem.

Nebius Token Factory is a managed inference and AI platform from Nebius for running open large language and image models in production. It provides premium open-source models, API integration, and infrastructure that scales automatically. The platform connects inference with production-data analysis, post-training, model optimization, and deployment. Its serving approach includes hardware-aware inference such as NVFP4, serving-engine selection, cache-aware routing, speculative decoding with custom draft models, KV-cache offloading, disaggregated prefill and decode, and quantization trade-offs, while handling infrastructure and serving optimizations.

Akamai LKE is the Kubernetes environment used in the workshop to run inference workloads on dedicated GPU nodes. It provides the namespace and deployment structure for a Jupyter notebook and a vLLM endpoint, and supports changing model-serving deployments and evaluating their latency, throughput, memory use, and output quality.

NVIDIA Dynamo is a developer toolkit from NVIDIA for managing distributed key–value (KV) context caches used in large‑language‑model inference. It provides KV‑aware routing, cache offloading, and prefill/decode disaggregation to move KV cache and related activation/state across GPU clusters, reducing memory pressure and supporting scalable LLM serving. It is intended for integration into GPU‑based inference and serving stacks.