What do you do?

We started by addressing the problem of model management: how to track, version, and audit models used across products, based on our research at MIT CSAIL building ModelDB – an open-source model management system deployed at multiple Fortune 500 companies – to create the Verta platform. Today, Verta provides model management and operations solutions for the entire AI & ML model lifecycle, from experiment tracking and production registry to deployment, inference & serving, and monitoring.

How much does it cost?

Enterprise pricing delivered as SaaS, on-prem, or VPC.

Feature List

  • Monitor model quality and performance: Ingest real-time and/or historical ground-truth and track model performance metrics like accuracy, precision, recall, f1, MSE, MAE etc
  • Detect drift: Track data drifts for input features and predictions and choose from a list of easily pluggable drift detection algorithms
  • Monitor outliers: Detect outliers and anomalous data for model features and predictions
  • Automated monitoring and alert creation: The system automatically configures features to monitor, alerts and thresholds with option to customize using an easy to use user interface.
  • Configurable metrics and dashboard: With a rich toolkit of charts, panels and widgets configure custom metrics and dashboards, build your own charts and visualizations (e.g. confusion matrix, PR curve, ROC curve and more)
  • Compare across release versions: Track and compare model performance and quality across models release and identify unexpected behavior
  • Root cause analysis: Connect pre-production (model registry and experiment tracking) and production systems for end to end visibility
  • Close the loop by fast recovery: Receive actionable alerts for performance degradation, or drift and automate remedial action like fallback, model retrain, human in the loop
  • Where can Verta be deployed ?

    Verta is a Kubernetes-based platform. Itcan deployed on any Kubernetes cluster, whether in the cloud or on-premise. Verta is offered as a hosted service managed by Verta or an enterprise deployment on-premise or in a VPC.

  • Which platforms and tools do you integrate with ?

    Verta platform is agnostic to any ML framework and Data prep/ML training platforms and integrates with them using a lightweight client library. This includes: Tensorflow, Pytorch, Sklearn, XGBoost, R, Spark, ONNX, PMML, Kubeflow, Anaconda, Spark, Sagemaker, AWS, VMware, Azure, Google CloudAdditionally we integrate with all the standard DevOps tools within the enterprise ecosystem. For example Github, PyPI, Jenkins, Kafka, Datadog etc.

  • Do you support SSO and work with enterprise identity providers ?

    The platform supports a robust RBAC (Role Based Access Control) system and integrates into enterprise identity systems (e.g Okta, Active Directory, etc.) along with the ability to define roles and permissions, and to audit activity on the platform. You can set up SSO (Single Sign On) and automate user provisioning with your enterprise identity provider.

  • Do you have access to our data ?

    We have different levels of deployment options to choose from starting from a fully managed SaaS service to fully air-gapped deployment. In the middle we also have a managed service option where we can push updates and manage deployment and have access to operational metrics only to provide 24X7 support without any access to your model at all.

  • Where can I find more information and examples on how to use the platform ?

    Given below are some of the relevant links –
    • Docs – User documentation with quickstart steps, concepts and guides for all code modules
    • Examples: In this repository you’ll find examples for the different modules of the Verta platform along with end-to-end examples spanning all parts of the platform
    API docs – API documentation of our client library


There are no reviews yet. Be the first to write one.