A real-world engineering story about building embedded analytics on Databricks - from Lakehouse Federation to AI/BI dashboards, tenant isolation, Delta materialization, refresh strategy, cost, and production lessons.

What does it actually take to build embedded analytics on Databricks for a production B2B SaaS application?

In this session, Andrii Taran, Lead Software Engineer at TechFabric, will walk through a real implementation that started with a six-line customer ticket and evolved into a production architecture spanning Lakehouse Federation, Unity Catalog, AI/BI Dashboards, Delta tables, tenant isolation, and embedded analytics.

This is not a “best practices” presentation. It’s a field report about what we built, measured, broke, rebuilt, and learned along the way.

We’ll cover how the team:

• started with Lakehouse Federation to get live application data into Databricks without building a pipeline;

• reduced dashboard query overhead and consolidated multiple reports into pages;

• implemented tenant isolation for embedded users;

• moved from federated queries to Delta materialization and improved dashboard performance by 5.4–6.7×;

• designed refresh schedules around actual business activity and cost;

• built a real Refresh flow for embedded dashboards;

• measured serverless cold starts, caching, and idle costs;

• handled Unity Catalog ownership and deployment pitfalls.

The session is aimed at data engineers, software engineers, architects, and anyone building customer-facing analytics on Databricks.

Speaker: Andrii Taran, Lead Software Engineer at TechFabric
Hosted by: Dnipro Databricks User Group
Organised by: TechFabric LLC with Databricks