It is 2 a.m. somewhere and a data pipeline just failed. Unravel Data built the thing that wakes up, finds the cause, and fixes it - so nobody else has to.
THE SUBJECT. Unravel Data's branded mark. A company that decided dashboards full of red alerts were a confession of failure, not a feature - and went looking for the off switch.
Walk into any large data team today and you will find a wall of dashboards. Green turning yellow turning red. A Slack channel that never sleeps. An engineer named, statistically, something like Priya, who can tell you exactly which Spark job blew the cloud budget last Tuesday - because she spent her afternoon finding out by hand. The modern data stack is a marvel. It is also, on a bad day, a very expensive smoke alarm: loud, certain, and completely incapable of putting out the fire.
Unravel Data was built by people who found this arrangement faintly absurd. Their argument is simple enough to fit on a napkin: a tool that tells you what is wrong, but cannot fix it, has only done half the job - and arguably the easier half. Anyone can point at a problem. Solving it at 2 a.m., across thousands of jobs, without breaking three other things in the process, is the part that actually matters.
"Your data platform should run itself. We're the company making that real."
So the company set out to build not another smoke alarm, but something closer to an autopilot. Its platform plugs into the systems where the world's data actually lives - Databricks, Snowflake, Google BigQuery, Cloudera - and watches everything: queries, jobs, infrastructure, datasets, the humans pushing buttons. It assembles all of it into what the company calls a context graph, a map of how every moving part relates to every other. And then, crucially, it does something with that map.
For most of its life, Unravel called itself a DataOps observability platform. That is industry-speak for "we help you see what is happening." It was good at it. The brains, fittingly, came partly from academia - co-founder and CTO Shivnath Babu spent over a decade studying database systems and taught the subject at Duke University. Co-founder and CEO Kunal Agarwal arrived from the other side of the table, having sold and implemented enterprise software at Fortune 100 companies and helped Sun Microsystems run its grid computing engine. One founder knew how data systems think. The other knew why enterprises lose sleep over them. The third, Paul Baclace, helped build the thing.
That pairing explains a lot. Unravel never treated observability as a pretty chart for its own sake. It treated it as the first step toward doing something useful - cutting a cloud bill, rescuing a failing pipeline, telling a finance team where the money actually went. The company leaned hard into FinOps, the unglamorous discipline of knowing what your cloud data habit really costs, complete with chargeback, showback, and forecasting. Less philosophy, more invoice.
Then artificial intelligence stopped being a slide in the pitch deck and started being the product. In 2024 the company shipped what it billed as the data industry's first purpose-built autonomous AI agents: a DataOps Agent, a FinOps Agent, and a Data Engineering Agent, each pointed at a specific, miserable, recurring chore. Not a chatbot bolted onto a dashboard - agents with a job description.
The logical conclusion arrived in May 2026 with Arvix AI, an agentic engine that does not merely suggest. It analyzes a workload, rewrites the code, reconfigures the infrastructure, and - this is the part that lets enterprises sleep - validates the change against existing work before it ships. If performance degrades afterward, it can quietly revert. Automation follows rules. Autonomy goes looking for trouble. Unravel bet on the second one.
One AI-native view across Databricks, Snowflake, BigQuery, Amazon EMR and Cloudera. Machine learning plus a context graph maps how every workload, dataset and user connects - so root cause stops being a scavenger hunt.
The autonomous optimization engine rewrites code, tunes infrastructure, validates changes before deployment, and reverts if things go sideways. You set the autonomy level. It does the night shift. Included in the platform, no separate license.
Cost visibility, forecasting, chargeback and showback for cloud data platforms. Turn a mysterious five-figure bill into a line item someone can actually govern - and a Snowflake spend you can defend.
Purpose-built DataOps, FinOps and Data Engineering agents that perform targeted tasks using domain knowledge graphs. Less firefighting, more building - which is, presumably, the job you were hired to do.
Roughly $119.5M raised across five rounds, capped by a $50M Series D in September 2022.
Revenue and headcount are third-party estimates. Bars are scaled for readability, not to a shared dollar axis.
Sold and implemented enterprise software at Fortune 100 firms, helped Sun Microsystems run its grid computing engine, and earlier co-founded e-commerce startup Yuuze.com. Knows why enterprises lose sleep.
Database systems researcher and former adjunct professor at Duke University. Spent over a decade studying the guts of data infrastructure - the platform's technical conscience.
The builder of the trio, with a hand in engineering and DevOps - turning the research and the sales instinct into something that actually ships and runs.
Large enterprises in finance, retail, pharma and tech - the kind of organizations where a runaway query is a budget line, not a footnote.
Agarwal, Babu and Baclace set up shop to make sense of big-data systems that were powerful but maddening to operate.
Named a 2018 Gartner Cool Vendor in Data Management - early validation that the diagnosis-plus-action thesis had legs.
Recognized by CRN in 2019, 2021 and 2022 as one of the coolest cloud companies.
Led by Third Point Ventures with Bridge Bank and existing backers Menlo Ventures, GGV Capital and Point72 Ventures, to accelerate the next generation of DataOps observability.
Launched purpose-built DataOps, FinOps and Data Engineering agents - the pivot from watching to doing.
Unveiled an agentic autonomous optimization engine that tunes and remediates Databricks, Snowflake and BigQuery workloads - rewriting code, validating, and reverting on its own.