Polyaxon

Polyaxon is an MLOps platform designed to help teams reproduce, automate and scale machine learning workloads.

Key features

• Job orchestration and scheduling: container-native DAG/workflow engine supporting parallel and distributed training.
• Experiment tracking and comparison: centralized logging of metrics and resource usage with dashboards and comparison views.
• Automation and hyperparameter tuning: built-in grid search, random search, Hyperband and Bayesian optimization.

Use cases

• Large-scale distributed training and hyperparameter optimization.
• CI/CD driven training pipelines and reproducible experiments.
• Multi-tenant resource sharing and team-level experiment management.

Technical notes

• Flexible deployment: self-hosted (Kubernetes/Helm), cloud-hosted or Polyaxon-managed services.
• CLI and SDK: polyaxon CLI, polyaxonfile configurations and SDKs for integration and automation.
• Modular architecture: submodules and plugins (e.g., hypertune, traceml) to extend functionality.