Databricks unveils Genie Code and buys Quotient AI

Databricks has launched Genie Code, an autonomous AI agent designed to handle multi-step data engineering, data science, and analytics work. It has also acquired Quotient AI as it pushes deeper into evaluating AI agents in production.

Genie Code sits within Databricks' Genie product line, which is positioned as a conversational interface for enterprise data that draws context and semantics from Unity Catalog. Genie Code targets data professionals and aims to turn ideas into production systems across pipelines, models, and dashboards.

Agent-driven work

Databricks is positioning Genie Code as a shift from AI tools that assist with writing code to systems that can plan and execute tasks with less step-by-step input. Genie Code can build pipelines, debug failures, ship dashboards, and maintain production systems, while keeping humans involved in key decisions.

"Software development has shifted from code-assistance to full agentic engineering in the past six months," said Ali Ghodsi, co-founder and CEO of Databricks. "Genie Code brings this revolution to data teams. We're moving from a world where data professionals are assisted by AI to one where AI agents do the work, guided by humans. We are calling this Agentic Data Work. It will fundamentally change how enterprises make decisions."

Databricks argues that existing agentic coding tools struggle with data tasks because they lack context such as lineage, usage patterns, and business semantics. It is relying on Unity Catalog to provide governance policies, access controls, and metadata for agents writing and changing code and interacting with datasets.

How it works

Genie Code is designed to run end-to-end workflows that span engineering and modelling. The agent can plan a workflow, write code, and validate results. It logs experiments to MLflow and can adjust serving endpoints.

It also targets common gaps between test and production environments. Genie Code is designed to account for staging and production requirements when it builds workflows, including change data capture patterns and data quality expectations.

Ongoing maintenance is another focus. Genie Code can monitor Lakeflow pipelines and AI models, triage failures, and investigate anomalies. It can also analyse agent traces to address hallucinations and tune resource allocation before a person intervenes.

Databricks is also highlighting persistent memory as a differentiator. Genie Code updates internal instructions based on prior interactions and coding preferences, adapting to how a team works over time.

On benchmarking, Databricks said Genie Code achieved a higher success rate than other coding agents on "real-world data science tasks," more than doubling from 32.1% to 77.1%.

Customer claims

SiriusXM is among the early users cited by Databricks. The media company said Genie Code spans tasks from notebook authoring to debugging pipelines.

"At SiriusXM, Genie Code supports everything from authoring notebooks and complex SQL to reasoning through table relationships and debugging pipelines," said Bernie Graham, VP of Data Engineering at SiriusXM. "It acts as a hands-on development partner that helps our data teams deliver high-quality work in less time."

Repsol also highlighted the role of governance and internal tooling, along with a broader set of workflows.

"Genie Code changes how our data teams operate," said Emilio Martín Gallardo, principal data scientist, Data Management & Analytics, at Repsol.

"Instead of stitching together notebooks, pipelines, and models manually, we can hand off complex workflows to an AI partner that understands our data, governance, business context, and internal libraries such as Repsol Artificial Intelligence Products. It accelerates everything from time series forecasting to production deployment, without sacrificing rigor or control," Martín Gallardo said.

Quotient acquisition

Alongside the product launch, Databricks has acquired Quotient AI, which it described as focused on evaluation and reinforcement learning for AI agents. The deal supports efforts to measure how agents perform once deployed and to manage regressions and failures over time.

Quotient's technology monitors agent performance and measures answer quality, flagging regressions early and identifying failure points. Databricks said this feeds a reinforcement learning loop to improve agents over time.

Quotient's founders previously worked on quality improvement for GitHub Copilot, according to Databricks. Databricks plans to embed the technology into Genie and Genie Code, and to strengthen Agent Bricks, another product in its portfolio.

Databricks said Genie Code integrates with Unity Catalog and can work across enterprise data, including data held on external platforms, with governance and audit requirements enforced through existing catalogue policies and access controls.

The rollout increases emphasis on autonomous execution inside Databricks' platform as enterprises test how far AI agents can be trusted with production changes. Databricks said the combination of catalog-driven context and continuous evaluation will shape how Genie Code operates as organisations adopt agents for day-to-day data work.