Unravel Data Systems Inc. is expanding beyond observability and FinOps software with a new autonomous optimization engine designed to automatically tune and remediate enterprise data platforms running on Databricks Inc., Snowflake Inc. and Google LLC BigQuery platforms.

The Palo Alto-based company today unveiled Arvix AI, an agentic artificial intelligence system embedded within the Unravel platform that analyzes workloads, rewrites code, optimizes infrastructure configurations and validates changes before deployment. The company said the technology reflects a broader industry shift from recommendation engines and dashboards toward autonomous operational systems.

“When you have humans making changes, you can only do so many fixes or improvements on these apps and platforms every day,” said co-founder and Chief Executive Kunal Agarwal. “Arvix AI is an agentic system inside of Unravel that builds off of decades worth of telemetry data to turn views and insights into action.”

Unravel Data noted that enterprises face escalating costs associated with cloud-based data and AI infrastructure, with complexity magnified by the need to run workloads across multiple platforms such as Databricks, Snowflake and BigQuery, while also expanding access to analytics beyond engineering teams.

“It’s no longer just engineers, but marketing, product, legal and finance teams on top of these systems,” Agarwal said. “That means more AI workloads that have to be checked for efficiency so they’re not wasting budget cycles.”

Service providers offer recommendation engines, such as Databricks Advisor and BigQuery Recommender, that can identify potential issues, but they still require engineers to implement fixes manually, Agarwal said. With Arvix AI, “you define a goal such as making sure all your data pipelines are running like well-oiled machines, and it just does that for you.”

He also drew a distinction between automation and autonomy, noting that conventional automation relies on deterministic, predefined actions, whereas Arvix AI continuously investigates systems and independently seeks optimization opportunities.

“It’s hunting for information by itself and figuring out the plausible reasons why something could be amiss,” he said.

Central to the platform is what Unravel calls a “context graph,” which maps relationships among workloads, infrastructure, data sets, applications and users across an enterprise. Agarwal said the decade of telemetry data the company has gathered enables it to model dependencies that are too complex for generic large language models.

The company emphasizes governance and safety features as critical requirements for enterprise adoption. Arvix AI validates proposed optimizations against existing workloads before deployment and continuously monitors the impact of changes after implementation.

“The last thing you want to do is change code and get the wrong answers,” Agarwal said. “It only makes a change if the new code is better.” The system can automatically revert modifications or reduce configuration changes incrementally if performance degrades.

Enterprises can determine what level of autonomy to grant based on workload characteristics such as service-level commitments or business criticality.

Agarwal said the industry is moving beyond dashboards and observability interfaces toward systems that directly execute remediation tasks. “We’ve been on this mission to change the word ‘observability’ to ‘actionability,’” he said.

Arvix AI is included in Unravel Data’s existing platform and won’t carry a separate license fee.

Image: Unsplash