Skip to main content

Getting Started with Built on Databricks

The Built on Databricks Partner Program provides access to technical and go-to-market resources to accelerate the development of your modern SaaS application and help you grow your business. Building on the Databricks Data Intelligence Platform provides a cost-effective, unified experience for developing data applications, products and services and enables seamless sharing of data at scale with a global open ecosystem.

Apply to the Built on Databricks Partner Program →

Reference Implementation​

See how Firefly Analytics implements the Partner Well Architected Framework patterns in a production reference implementation. Firefly demonstrates SSO-SPN authentication, workspace isolation, cost management, and all the architectural patterns documented in this guide.

Validation​

Built on Databricks validation certifies that your product is built on the Databricks Data Intelligence Platform and that its architecture meets the Partner Well Architected Framework across our core principles: Governance, Cost Management, Scale and Limits, Automation, and Onboarding.

Process​

PhaseWhat happens
ApplySign up for the Built on Databricks Partner Program at partners.databricks.com
DevelopBuild your architecture following Partner Well Architected Framework best practices, using the Firefly reference implementation as a model
SubmitWork with your Databricks account team to submit your application
ValidationThe Databricks Partner Engineering team works with you to validate your architecture
GTM OnboardingOnce validated, the Partner Operations team works with you on go-to-market assets

Requirements​

To be validated, your product must meet the following requirements, and you must submit the assets below for review.

Technical requirements​

  • Run on Databricks: Databricks is the core data and AI platform your product is built on, not a system it connects to
  • Use a defined deployment and tenancy model: a clear deployment model and tenancy model with appropriate per-customer isolation
  • Govern data in Unity Catalog: all production data assets governed by Unity Catalog (managed tables, access control, lineage, auditing)
  • Authenticate securely: OAuth M2M (service principals) for backend access and SSO-mapped / U2M OAuth for end users; prefer Workload Identity Federation; no Personal Access Tokens
  • Attribute cost per customer: resource tagging so Databricks cost can be attributed to each customer
  • Separate product from internal consumption: the Databricks consumption that powers your product is segregated from your own internal/corporate usage, so product-level consumption can be measured
  • Attribute Genie: if your product surfaces Genie results, display "Powered by Genie" and cite the source Genie Space, per the Genie attribution requirements
  • Use the Genie MCP: if your product is an AI agent or chat-based interface, integrate with Databricks via the Genie MCP

Required assets for validation​

  • Reference architecture diagram, defining the elements listed below
  • Recorded product demo: a video walkthrough of your platform that showcases the architecture and the customer experience
  • Databricks account and workspace details: the account ID(s) and workspace(s) where your product runs, so the Partner Engineering team can verify Unity Catalog adoption, cost tagging, and authentication

Your architecture diagram should clearly show​

  • Deployment pattern: e.g. Partner Hosted SaaS, Customer Managed, Hybrid
  • Tenancy and isolation: how each customer is isolated (separate workspaces, Unity Catalog catalog/schema per tenant, row-level security, separate storage)
  • Customer access: embedded through your application vs. logging directly into Databricks
  • Databricks resources and products used
  • Authentication: both end-user and backend
  • Data flow: how data is ingested, stored, processed, and delivered to customers (e.g. Delta Sharing, APIs, file export)
  • AI/ML components: which you use and where they run (Databricks Model Serving, Foundation Model APIs, Agent Bricks, Vector Search vs. external providers)
  • Components off Databricks: proprietary engines, transactional databases, BI layers, external model providers
  • Product scope: which product this validation covers, if you offer several

Validated partners are subject to revalidation as needed when Databricks updates its validation requirements, and must comply on an ongoing basis with the Databricks Partner Program Terms and the Databricks Acceptable Use Policy.

Additional Resources​