Third-Party Integrations
This section covers connecting third-party tools to Michelangelo AI. If you are looking for documentation on Michelangelo AI's own components — the model registry, experiment tracking setup, serving infrastructure, or job scheduler — see the Operator Guides index.
Before configuring any tool below, complete Experiment Tracking Setup — the platform-level guide for network reachability, ConfigMap injection, auth, and the operator/user boundary that applies to all third-party tracking integrations.
| Guide | Description |
|---|---|
| Comet ML | Connect to Comet ML's experiment tracking — network setup, API key injection, PyTorch Lightning / Ray Train / HuggingFace Transformers / custom training loop patterns, distributed experiment coordination, and Comet ML vs Michelangelo AI model registry comparison |
| MLflow | Connect a self-hosted or Databricks-managed MLflow Tracking Server — network setup, auth, and MLflow vs Michelangelo AI registry comparison |
Next Steps
- Network & Ingress — configure egress rules, Envoy proxy, ingress, TLS, and multi-cluster networking
- Authentication — manage secrets, workload identity, and RBAC for credential handling
- Troubleshooting — diagnose common failure modes with
kubectldiagnostic commands