Seldon

The Seldon mission is to accelerate the adoption of ML to improve business performance and manage risk. Backed by this mission, Seldon is building the global infrastructure for ML operations.

Seldon is backing and offers enterprise support for the below products.

Deploy Advanced is the proven platform to serve, monitor, explain and manage your machine learning models. Seldon Deploy Advanced enables the orchestration of models and monitoring components at scale.

• Workflow and Configuration Wizards for simple and faster setup and deployments.
• Canary and A/B Testing for constant optimisation and peak performance.
• Enterprise APIs with SDK resulting in huge time savings and enhanced collaboration across the business.
• Configurable Multi-Model Serving allows you to deploy and host multiple models more cost-effectively.

Advanced observability & monitoring can be added to your running deployments with built-in dashboards for distribution, drift and outlier detection. Set up alerts when unexpected behavior occurs and take action on production data issues in real time.

• Operational Dashboards to view performance of all deployed assets in one place.
• Drift Detectors so you can detect shifts in a model’s data dynamics in order to understand when to retrain models across all data modalities – be it text, tabular, image, time series or categorical.
• Outlier Detectors to flag when there are anomalies in model predictions to decide if it is erroneous enough to retrain models across all data types.
• Error Alerting allows you to quickly respond to incidents  before any negative impact on the business. 

Advanced observability & monitoring can be added to your running deployments with built-in dashboards for distribution, drift and outlier detection. Set up alerts when unexpected behavior occurs and take action on production data issues in real time.

• Operational Dashboards to view performance of all deployed assets in one place.
• Drift Detectors so you can detect shifts in a model’s data dynamics in order to understand when to retrain models across all data modalities – be it text, tabular, image, time series or categorical.
• Outlier Detectors to flag when there are anomalies in model predictions to decide if it is erroneous enough to retrain models across all data types.
• Error Alerting allows you to quickly respond to incidents  before any negative impact on the business. 

Intuitive logging, alerting, and audit trails allow for clearer understanding of ML pipelines across teams. Gain enhanced user management for granular policies and governance.

• Open Inference Protocol improving portability to platforms that also use this API as we drive for more standardization in the ML industry. 
• Audit Logs using GitOps facilitates reproducibility, the ability to revert back to a previous state of your platform when something goes wrong and when coupled with user management, know exactly who accessed your system to make changes. 
• Model Catalog offers an overview of all metadata related to your model from a single repository including custom fields to help you track model information more effectively. 
• Granular User Management allows you to apply specific policies to each user or user group to define access rights down to the model-level so you can control which team members can create, view, update or delete assets in your environment. 

Seldon IQ is an add-on to Seldon Deploy Advanced that enables customers and partners to achieve exponential results with the suite of Seldon products through enablement, workshops and certifications.

Seldon Core is the open-source framework for easily and quickly deploying models and experiments at scale. Seldon Core enables data-centricity and easier model deployments through experiments, ensembles, and transformers. Runs on any cloud, is framework agnostic, and supports top ML libraries, toolkits, and languages. Seldon is backed by a thriving MLOps community with over 3,000 members.

Seldon Core handles scaling to thousands of production machine learning models and provides advanced machine learning capabilities out of the box including Advanced Metrics, Request Logging, Explainers, Outlier Detectors, A/B Tests, Canaries and more.

Open source Python library aimed at machine learning model inspection and interpretation. When your data changes so can your models predictions. Ensure accuracy by monitoring alteration. Gain insights into how features influence model performance. Return the score that indicates the presence of features & instances that trick the model outcome.

Open source Python library focused on outlier, adversarial and drift detection. Notice changes in data dynamics & define whether detected drift will cause a decrease in model performance. Discover critical anomalies in input and output data using outlier detection.

Alert business units and users when seeing unexpected behavior. Adversarial detection ensures that models perform consistently. Return the score that indicates the presence of features & instances that trick the model outcome.