Seldon

What is Seldon?

Seldon is a Kubernetes-native MLOps company whose Seldon Core 2 framework deploys, manages and scales traditional ML models and LLM applications. Its MLServer inference server is open source, and REST and gRPC inference follows the Open Inference Protocol.

Key capabilities of Seldon

• Kubernetes model serving: deploys models at scale
• MLServer: open-source inference server for multiple frameworks
• Monitoring and drift: observes production models
• Explainability: explains predictions via Alibi
• LLM deployment: supports LLM applications
• Multi-model serving: runs many models efficiently

How Seldon works

Teams define models and pipelines in Kubernetes and Seldon Core 2 serves them through REST or gRPC endpoints that follow the Open Inference Protocol. MLServer runs the models, with multi-model serving. Monitoring and explainability components help watch drift and understand predictions. Seldon Core 2 was released in January 2024 under the Business Source License.

Who uses Seldon?

ML platform and MLOps teams running models on Kubernetes use it. Enterprises can choose Seldon Enterprise Platform or Core+ for support and unlimited models and users.

Seldon pricing

MLServer is free under Apache 2.0, and Seldon Core 2 is source-available under BSL 1.1 that converts to Apache 2.0 four years after each version. Seldon Enterprise Platform and Core+ use custom annual pricing. Third-party sources report these figures.

Seldon alternatives

Alternatives include ZenML for pipeline orchestration, Evidently AI for monitoring and drift, Amazon SageMaker for managed ML on AWS, Vertex AI for managed ML on Google Cloud, and Hugging Face for model hosting.