ml-infrastructure

To deploy a machine learning model to Kubernetes, you package the model into a Docker container and use Kubernetes manifests to scale the serving endpoints. This skill scaffolds the containerization and deployment configuration for production model serving.

Building an ML pipeline with MLflow involves logging preprocessing steps, training metrics, and model versions during orchestration. This skill scaffolds reproducible training workflows that integrate MLflow tracking directly into the pipeline.

Monitoring machine learning models for drift requires exposing metrics like latency and data distribution changes. This skill provides guidance on setting up Prometheus endpoints to validate production models and trigger retraining workflows.

Yes, you can use FastAPI to serve machine learning models by containerizing the endpoint and deploying it on Kubernetes. This skill provides artifacts for packaging FastAPI servers and scaling them using Kubernetes manifests.

Yes, this MLOps skill supports orchestration with Kubeflow and Vertex AI. It provides guidance for end-to-end pipeline orchestration, from data preprocessing to model training and deployment using these platforms.

Integrating a feature store into an ML training pipeline requires connecting preprocessing outputs to the store and retrieving features during orchestration. This skill provides guidance for feature store integration within scalable training workflows.