seldon core
**Seldon Core: Kubernetes ML Deployment**
**Overview**
Seldon Core is an MLOps framework specifically designed to deploy machine learning models on **Kubernetes**. It converts your model into a production-ready microservice with metrics, logging, and scaling.
**Key Features**
**1. Inference Graphs**
You can chain models together.
- Input -> [Preprocessing Model] -> [Classifier A] -> Output.
- Input -> [Router] -> (Model A or Model B) -> Output (A/B Testing).
**2. GitOps Friendly**
You define your deployment as a Kubernetes YAML manifesto.
```yaml
apiVersion: machinelearning.seldon.io/v1
kind: SeldonDeployment
metadata:
name: sklearn
spec:
predictors:
- graph:
name: classifier
implementation: SKLEARN_SERVER
modelUri: s3://my-bucket/model
```
**3. Standard Metrics**
Automatically exports request count, latency, and custom metrics to Prometheus/Grafana.
**4. Explanations**
Native integration with **Alibi** (Explainable AI library) to explain *why* the model made a prediction.
**Use Case**
Seldon is "Heavy Duty". Use it if you are already running Kubernetes and need to manage hundreds of models at scale in an enterprise environment.