Speed up the turnaround time of AI results

Deployment Modes

When introducing a new model or a new version of the model, operations teams may want to test with live traffic and compare the results with the previous version. With MLOps, users can run comparison tests between the production model (champion) and an experimental model (challenger), or test two (or more) models and compare the results in a simple A/B test.

Various environments

MLOps uses customer-provisioned Kubernetes environments and supports multiple infrastructure environments simultaneously. MLOps teams can have development, test, and production environments all running in different locations.

Deployment Type

Easily service your model as a real-time deployment (synchronous or asynchronous) or batch deployment (one-time or scheduled) with the click of a button.

Updates and Reversals

Machine learning models can require frequent updates in production. With MLOps, operations can easily replace templates with just a few clicks, and Kubernetes automatically handles routing new requests to the new template, while the previous version handles old requests. Likewise, if there is a need to roll back to a previous version, this can also be done with the click of a button.

Model Drift

When data changes between training and production, models can become less effective. This "drift" is tracked by looking at the differences between training and production data for each feature in the model. We offer a deviation detection AI application designed for data scientists, which has detailed visualizations for each feature so data scientists can determine whether to readjust, retrain, or create a new production model.

Accuracy

As models are in production for some time, model performance and accuracy generally degrade. For organizations to have confidence that their models are predicting as desired, model accuracy must be monitored. We have monitoring capabilities for regression models and classification models.

Justice and trend

Machine learning models are inherently subject to biases, which can be caused by various factors. Fairness is not only an ethical issue, but also fundamental to achieving optimal business value. MLOps monitors models for biases to provide organizations with information to derive better business value from their models.

Operational

Production models are typically served on infrastructures prone to operational problems. Sudden spikes in request volume or problems with model objects can cause increased latency or, even worse, deployment disruption. MLOps monitors organizations' operational metrics, so IT teams can detect and respond to problems before they become business issues.

Custom Limits

For each deployment, users can set thresholds and alerts on a variety of metrics. When metrics reach the given point, an alert is triggered to notify the MLOps team about the problem so they can take appropriate action.

User and Group Permissions

Set permissions for users and groups for Projects, enabling the appropriate access level.

Artifact Sharing

Share templates and template artifacts with collaborators and team members and give them permission to perform actions for their templates.

Analysis Panel

Receive a complete view of all models across the organization and gain insight into how models are progressing in the deployment workflow. Build a deeper understanding of machine learning adoption within the organization by mapping models to their creators.

Governance, Compliance, Accountability

Data, Experience, Model, Deployment Lineage MLOps maintains data and traceability throughout the machine learning lifecycle, enabling end-to-end data lineage, experimenting, modeling the record, and deployment.

Reproducibility

With each template and deployment maintaining its metadata, organizations can replicate template development for any type of compliance need.

Explanations of the model in production

Receive model explanations for each scoring request at runtime. Model explanations make it easy for individuals to understand which AI model features contributed most, both positively and negatively, to each individual prediction. With run-time explanations, organizations can more easily validate, analyze, and improve model results while being compliant with industry regulations.

Event log per deployment

X'MLOps includes an event log for each deployment. The log captures all events related to the deployment, including who performed the action and when it occurred.

High-availability deployments

Select up to 5 nodes to replicate template deployments. MLOps will automatically check the health of each node and load balance between nodes, so if a node fails, clients will not experience any service interruption.

Configuring Deployment Infrastructure

Choose the Kubernetes configuration best suited for each model, including CPU vs GPU and minimum and maximum processing and memory allocations. Customers can assign deployments to GPU nodes when the model size is large and/or requires ultra-low latency scores.

Deployment options

MLOps, is our broad and expanding suite of products. Customers deploy and use the entire next-generation AI platform or just the products that meet their needs. There are two deployment options, hybrid and fully managed.