Databricks is a powerful environment for developing and serving machine learning models but as organizations scale, governance, transparency, and compliance become harder to maintain across teams and tools.
With our latest Databricks integration, Deeploy makes it seamless to bring all your models, whether externally hosted or natively cataloged, into a single, centralized governance layer. That means full control, oversight, transparency and compliance tracking without interrupting your current workflows.
We now support two complementary integration paths:
Onboard existing Databricks models into Deeploy
With our new bulk onboarding capability, you can now connect Databricks serving endpoints as external Deployments in Deeploy.
This lets you keep your models hosted and running in Databricks, while layering on Deeploy’s powerful governance, monitoring, and compliance capabilities.
How it works:
- Head to the Integrations tab in Deeploy and configure your Databricks credentials.
- Select the serving endpoints you want to onboard.
- Deeploy automatically creates corresponding external Deployments for each endpoint.
Your models remain hosted in Databricks, but are now fully visible and manageable in Deeploy. From this point, you can:
- Monitor model performance and usage
- Track and log predictions
- Assign model owners and roles
- Attach documentation, checklists, and compliance records
- Set up real-time alerts
This is ideal for teams that want to maintain their current Databricks infrastructure but need to layer in centralized governance and oversight.
Deploy models directly from Databricks Unity Catalog
For organizations managing models in Unity Catalog, we now support deploying managed Deployments in Deeploy directly from your Databricks workspace.
This allows you to deploy models (and explainers) from your Databricks model registry into Deeploy using either a specific version or a model alias, no Git repository needed.
How it works:
- Set up your Databricks integration in Deeploy
- In the deployment flow, select your model from Unity Catalog
- Choose to deploy by version or alias
- Deeploy automatically generates the correct reference
Once deployed, you unlock the full capabilities of managed deployments:
- Monitor model performance and usage
- Track and log predictions
- Explain predictions to enhance model transparency
- Assign model owners and roles
- Attach documentation, checklists, and compliance records
- Set up real-time alerts
- Easily update with new model versions
Oversee and manage all models from one place
Bring all your Databricks models into a single, centralized governance layer with Deeploy.
You gain full control and oversight without disrupting existing workflows. With built-in monitoring, real-time alerts, and performance visibility, it’s easy to catch issues like model drift early and ensure your AI systems remain safe, transparent, and effective at scale.
Stay compliant and audit-ready
Deeploy makes compliance easier by auto logging all model outputs and decisions in one centralized place, so you’re always audit-ready.
You can also attach documentation, manage checklists, and organize and data cards, all without leaving your workflow.
Frequently asked questions
Does Databricks offer built-in AI governance capabilities?
Databricks provides powerful infrastructure for building and deploying ML models, but it does not offer dedicated AI governance features like audit trails, compliance tracking, explainability, or risk documentation. Deeploy fills this gap by integrating directly with Databricks, adding the controls needed to meet regulatory standards such as the EU AI Act or ISO 42001, without disrupting your existing workflows.
Does Deeploy change how models are hosted in Databricks?
Using external deployments, Deeploy connects to models already hosted in Databricks, allowing you to add governance without changing where the models run. However, certain features, such as explanation requests and container logs, are not supported in this setup.
With managed deployments, models are deployed to Deeploy via Unity Catalog and hosted on Deeploy’s infrastructure instead of Databricks. This shift enables full access to Deeploy’s functionality, including advanced monitoring, governance workflows, and model explainability, while still letting you version and manage models directly from within Databricks.
Can I explain predictions made by my Databricks models?
By hosting your models as managed deployments in Deeploy, you gain access to built-in explainability features. Techniques like SHAP and counterfactuals can be used to visualize and log prediction reasoning, helping you meet transparency and accountability requirements directly within the platform.
How does this help with AI compliance (e.g., EU AI Act, ISO 42001)?
Deeploy tracks model usage, performance, ownership, and documentation, all in one place. It supports control frameworks, audit logs, human-in-the-loop decisions, and alerting, all key capabilities for demonstrating compliance with key AI regulations such as the EU AI Act.
