dstack

dstack is an open-source orchestration layer for running AI workloads across GPU clouds, Kubernetes, and on-premises infrastructure using different hardware and model frameworks. Teams should secure credentials and images, restrict network access, pin versions, validate cost and capacity, test failover, monitor workloads, review licenses, and protect training data and model artifacts.

Start with authorized, minimal, non-sensitive inputs. Configure privacy, retention, permissions, quality, disclosure, export, moderation, security, and spending controls; compare results with authoritative sources and real requirements; correct errors; test integrations and automations in a limited environment; and retain accountable human approval before sending, publishing, buying, deploying, or making consequential changes.

The dstack orchestration software is open source and free to self-host. GPU cloud, Kubernetes, storage, networking, managed infrastructure, enterprise support, and third-party provider charges remain separate.

AI output can be inaccurate, generic, biased, incomplete, stale, unsafe, insecure, or misleading. Review privacy, training, retention, copyright, consent, commercial rights, renewals, refunds, integrations, and professional limits. Education, surveillance, safety, finance, legal, and other high-impact work requires qualified human review.