dstack is an open-source, unified control plane designed to simplify the orchestration of AI/ML infrastructure. It allows engineers and autonomous AI agents to provision and manage compute resources for training, inference, and development environments across various GPU types (NVIDIA, AMD, TPU, Tenstorrent) and platforms, including cloud, Kubernetes, and on-prem clusters. dstack offers fine-grained resource management, native support for container workloads, and easy integration with existing infrastructure, targeting both individuals and teams in ML/AI-focused organizations.
Visit Dstack's official website for product details and getting started.