Lakeflow Designer: No-code ETL— Authored With AI, Built for Production

The result is a visual canvas where non-coders transform, join, and visualize data, while every step compiles to clean PySpark that fits existing CI/CD, scheduling, and access controls. Work stays in one governed place, and pricing follows standard Databricks compute, not per-seat licenses.

Why a native no-code experience changes the workflow

External low-code tools like Alteryx, DataIQ, PowerQuery, Excel, or Google Sheets often pull data out of a governed platform, duplicate efforts across desktops, and create brittle handoffs for production. They also add yet another system for IT to manage and often impose high per-user costs.

Lakeflow Designer keeps preparation, sharing, and promotion to production inside the Databricks Platform. Analysts work visually, engineers inherit code that already conforms to platform operations, and governance remains intact from day one. Organizations can extend access broadly without rationing seats.

AI that is powerful, stepwise, and reviewable

Genie, the built-in natural language assistant, converts plain English requests into transformations and iterates in an agentic loop. It inspects inputs, proposes changes, validates outputs, and materializes each step as an operator on the canvas with side-by-side input and output samples.

A demo builds a World Cup analytics table by combining a Databricks table with an uploaded Excel file, parsing dates, filtering since 2012, aggregating, computing a custom power rating, and rounding to spec. The most striking moment is "transform by example": pasting a screenshot of the desired table and asking Genie to match it. The assistant infers joins, aggregations, formulas, and decimal precision, then renders the steps for human review and adjustment.

Reusable building blocks with user-defined operators

User-defined operators, or UDOs, let teams package reusable logic as drag-and-drop blocks. A team can, for example, enrich a dataset with latitude, longitude, continent, and ISO codes from a country field, or render a globe visualization, all without analysts writing code.

UDO definitions consist of a YAML blueprint for inputs, outputs, UI controls, and metadata, plus plain Python for the implementation. Because the code is standard Python, teams can unit test, version, and manage UDOs through normal development practices. Common uses include encapsulating KPIs and formulas, exposing statistics or data science functions to analysts, and standardizing API integrations. The speakers also show generating a simple custom operator on the fly from natural language, such as mapping ISO codes to flag emojis.

From canvas to production without a rewrite

Every action on the canvas compiles to plain PySpark. Pipelines can be stored in Git folders, parameterized, scheduled, included in Databricks asset bundles, and wired into CI/CD. Iteration remains fast, since operators default to running on data samples, and users can adjust sample size or switch to full data when needed. Visualization is built in for quick charts, and custom UDOs can add domain-specific visuals where helpful.

A practical bridge between analysis and operations

Lakeflow Designer shortens the distance between the people who shape the questions and the systems that run in production. Visual steps stay transparent, AI assistance remains reviewable, and the output is deployable code under existing governance. Teams gain a shared surface where analysts move faster and engineers keep operational control.