Seldon

Seldon is a machine learning deployment, monitoring, and explainability platform built for organizations that need production-grade model serving with robust governance capabilities. Founded in 2014 and headquartered in London, United Kingdom, Seldon has established itself as a leader in the ML deployment space, particularly among organizations running Kubernetes-based infrastructure. The company offers both open-source tools and an enterprise platform that together provide a comprehensive solution for deploying, managing, monitoring, and explaining ML models in production. Seldon's core offering is built around three main components: Seldon Core, an open-source Kubernetes-native model serving framework that supports deployment of models built with any ML framework; Seldon Deploy, an enterprise platform for model deployment management with dashboards, canary deployments, A/B testing, and multi-armed bandit optimization; and Alibi, an open-source library for ML model inspection and interpretation that provides algorithms for model explanations, drift detection, and outlier detection. The governance capabilities of Seldon are notably stronger than most MLOps platforms. Alibi Explain provides model-agnostic and model-specific explanation algorithms (SHAP, LIME, anchors, counterfactuals) that help organizations meet transparency and explainability requirements. Alibi Detect offers drift detection (data drift, concept drift, adversarial detection) and outlier detection that support ongoing model monitoring governance requirements. Seldon Deploy adds enterprise governance features including model approval workflows, audit trails, role-based access control, and deployment policies that enforce governance guardrails around model promotion to production. Seldon's Kubernetes-native architecture enables seamless scaling, versioned deployments, and traffic management for sophisticated deployment strategies like shadow deployments (testing new models against production traffic without serving predictions) and progressive rollouts. The platform integrates with MLflow, Kubeflow, Airflow, and major cloud platforms. However, Seldon's Kubernetes dependency creates a significant barrier to entry for organizations without Kubernetes expertise. The full enterprise feature set requires the paid Seldon Deploy product, and the open-source components alone may not provide sufficient governance for enterprise requirements. Setup and configuration complexity can be considerable, particularly for advanced deployment patterns and the explainability pipeline integration.