moai-domain-ml-ops

End-to-end MLOps orchestrates the full ML lifecycle using MLflow for experiment tracking, DVC for data versioning, Kubeflow for workflows, and Ray Serve or Seldon Core for model serving. This Skill integrates these tools into a cohesive Kubernetes-based platform that manages experiments, data, pipelines, and deployments together.

Yes. This Skill unifies MLflow, DVC, Kubeflow, and complementary tools like Feast for feature stores, Optuna for hyperparameter optimization, and Evidently AI for monitoring. It handles integration across all stages—experiment tracking, data versioning, orchestration, and serving—within Kubernetes deployments.

Before deployment, you need experiment tracking configured in MLflow, versioned datasets in DVC, orchestrated workflows in Kubeflow or Argo, and hyperparameter optimization through Optuna. This Skill assumes these prerequisites are in place and coordinates their outputs into a deployable model-serving pipeline.

Production monitoring uses Prometheus for metrics collection and Grafana for visualization, integrated with Evidently AI for model performance tracking. This Skill combines observability from artifact storage, model serving logs, and inference metrics into unified dashboards.

Yes. This Skill provides automated CI/CD within Kubernetes by orchestrating MLflow, DVC, Kubeflow, and Argo workflows. It handles continuous experiment tracking, data versioning, pipeline execution, and model deployment as integrated, repeatable stages.

Both Ray Serve and Seldon Core are supported for model serving in this platform. Ray Serve scales within Ray clusters; Seldon Core deploys natively on Kubernetes. This Skill accommodates either, depending on your cluster architecture and scaling requirements.