ACTIVATE is a control plane that connects researchers and engineers to computing resources on-premises, in the cloud, and through GPU providers, all under a single identity. Parallel Works offers ACTIVATE as a managed service, or customers can deploy it within their own cloud accounts or data centers.
One login for compute, data, workflows, and AI
Users reach ACTIVATE from the browser, the CLI and API, or AI assistants. The control plane handles identity and groups, provisioning, workflows, sessions, the AI gateway, and allocations and metering across:
- Cloud clusters on AWS, Azure, Google Cloud, and Oracle Cloud
- GPU providers, including CoreWeave, Lambda, and Lightning AI
- On-premises systems running Slurm and PBS
- Kubernetes clusters, with namespaces by group
- Storage, including buckets and filesystems
- AI models, hosted and external
One workflow definition, in YAML, runs across connected computing resources, and published workflows launch through automatically generated input forms. A consistent codebase, hardened containers, and standardized base images run on the shared Parallel Works platform, in a customer's cloud account, and on customer hardware.
Managed GPU clusters
Parallel Works runs a dedicated GPU reservation at a cloud or GPU provider as one managed cluster, with per-team accounts and quotas. The provider runs the GPU nodes, fabric, storage, and the data center. Parallel Works runs the platform and the cluster, with Tier 1 end-user and application support. Workloads can move between providers while keeping the same user identities, group mappings, Slurm accounts, and storage paths, so users are not onboarded again.
Cost, education, and licensing
Every cost record carries the user, group, and project, and budget thresholds set by an administrator trigger notifications, freeze allocations, or stop workloads. Universities connect campus single sign-on, extend workloads from campus clusters to cloud resources while maintaining consistent user access, and run notebooks, remote desktops, and VS Code sessions for courses and labs.
Teams of approximately 100 users or fewer are licensed on a per-user basis through the shared platform. Larger programs are licensed on a per-deployment basis and include unlimited users, connected resources, and support from a Parallel Works mission engineering team. Academic institutions receive a 50% discount on commercial list pricing. See pricing for details.