Skip to main content

Welcome to Anyscale

Bring your most ambitious AI workloads. Anyscale simplifies the ML infrastructure, in your cloud or ours.

Understand the basics​

Build and deploy modern machine learning systems on Anyscale's optimized compute platform, created by the team behind Ray. Scale generative AI workloads like embedding, fine-tuning, batch inference, serving, and other distributed workloads on managed compute.

How does it work?​

Anyscale-hosted cloud executes your code within a dedicated cloud environment that we manage and optimize on your behalf. You can get started quickly without engaging with Kubernetes, Docker, or cloud service providers.

For enterprises, self-hosted clouds run on your AWS or GCP accounts with Anyscale orchestration. See Architecture for more details.

Why Anyscale?​

Thousands of organizations use Ray to increase developer velocity and distribute ML and Python apps. Anyscale not only provides the best environment for developing with Ray, but also adds essential enterprise features:

Production-grade scalable compute platform​

  • Fully managed service: Anyscale handles cluster management, enabling ML practitioners to concentrate on coding scalable applications rather than infrastructure.
  • Bring your own cloud: Anyscale operates within your cloud account, offering a secure environment for customer data.
  • Cost optimization: Features like spot instance support, autoscaling, and auto-suspend in Anyscale reduce compute expenses, complementing your cloud provider agreements like AWS Reserved Instances and Savings Plans.
  • Governance and compliance: Anyscale ensures robust access controls and cost tracking, backed by SOC 2 Type 2 attestation.
  • Expert support: Receive direct support from the creators of Ray and a team of dedicated Anyscale engineers.

Unified development environment for ML and Python workloads​

  • All-in-one development environment: Anyscale Workspaces offers an IDE-centric experience for code development, dependency management, and observability. It feels like your laptop, but you have the power of a cluster.
  • Use the tools you love: Instant setup with popular tools like VS Code, Jupyter, GitHub, Weights & Biases, and more.
  • Collaborate: Instantly share or replicate experiments with the same compute and environment, allowing quick collaboration on your team.

Seamless transition between development and production​

  • Unified environment for development and production: Develop, run, debug, and test code at scale reusing compute configurations and container images between development and production.
  • Flexible dependency management: Manage dependencies with Anyscale base container images or your build own image.
  • Jobs and services: Simplified interface for production workloads. Anyscale Jobs support cron jobs, ephemeral clusters, and retries. Anyscale Services provide replica management, zero downtime upgrades, and high availability.
  • Comprehensive logs and monitoring: Managed Grafana and Ray Dashboard for robust observability and alerting.

Get support​

  • Official support: To reach out to the team directly, email support@anyscale.com.
  • Community channel: Join the Ray Slack and chat in the #hosted-anyscale-support channel.
Special requests

Reach out to us at support@anyscale.com if you want access to the following features on Anyscale-hosted clouds:

  • Access to A100 and H100 GPUs
  • Fully hosted cloud on Google Cloud Platform
  • Custom cloud storage for data retention
Storage retention policy

Anyscale provides temporary object storage as scratch storage. $ANYSCALE_ARTIFACT_STORAGE retains data for up to 90 days from the date of creation. Anyscale removes all user data within a maximum of 120 days following account deactivation. Anyscale-hosted clouds may retain logs for up to 2 years. Users can request deletion of their Anyscale-hosted cloud account by contacting support@anyscale.com.