Databricks Expands Genie Code

Databricks is evolving its AI assistant, Genie Code, with significant updates unveiled at the Data + AI Summit 2026. The company is pushing towards AI-native workflows with enhancements designed to streamline complex data and machine learning development.

Genie Code, the specialized agent for data and ML tasks on Databricks, is getting a new full-page command center. This dedicated space aims to simplify the management of intricate, multi-threaded work, offering visibility into thread status, review points, and direct access to instructions, skills, and connectors.

Managing Complex Workflows

Data and ML projects often involve more than a single prompt. The new command center addresses this by providing a consolidated workspace for tasks that span notebooks, SQL, pipelines, and various Databricks assets. Users can now manage multiple Genie Code threads concurrently, track their execution, and return to them as new results become available. This offers a more organized approach compared to managing tasks across numerous tabs.

The enhanced interface also improves discoverability of customization options. Instructions, skills, and connectors are more prominent, allowing teams to better guide Genie Code with specific standards and workspace knowledge.

Agentic Development for ML

Databricks is also enhancing Genie Code's capabilities for production ML engineering. The updates focus on the surrounding engineering tasks that consume significant time and resources in ML projects, such as feature engineering, experimentation, and deployment.

Genie Code now leverages Databricks' decade of experience in production ML and integrates with the team's specific patterns through Genie Ontology. This allows the agent to handle details like class imbalance correction and feature quality checks, mirroring the actions of a seasoned practitioner.

Key integrations include MLflow, where Genie Code can read experimentation and observability data to answer questions about training optimization or metrics. It also inspects Model Serving endpoints for health and performance issues, diagnosing problems and suggesting optimizations.

Furthermore, Genie Code gains compute awareness, automatically leveraging GPUs for training jobs and utilizing workspace environment features to simplify infrastructure setup. This deeper integration aims to enable data scientists to complete real-world tasks more efficiently.

Autonomous Tasks on the Horizon

Looking ahead, Databricks is introducing scheduled tasks for Genie Code. This feature will allow the AI agent to perform work autonomously, even when users are offline. Scheduled tasks will initiate a thread with results for user review, enabling data teams to automate tasks like checking job outcomes, summarizing pipeline runs, or preparing weekly analyses.

This move shifts Genie Code from an interactive assistant to a more autonomous worker. The company also highlighted Genie ZeroOps, which extends agentic automation into production operations, monitoring live systems and preparing fixes for review.

These advancements are part of a broader trend at Databricks toward AI-native data and ML workflows, aiming to accelerate the entire data lifecycle from development to operations. The company has seen significant growth in its Genie products, indicating strong adoption by customers.