Coalesce - the quiet plumbing behind your data stack.

Somewhere inside a Snowflake account at a Fortune 500 retailer, a data engineer presses a button. A pipeline that used to take eight weeks and four meetings now ships in an afternoon. Nobody throws a parade. The dashboard is just - correct. This is what Coalesce sells, and it sells it very deliberately quietly.

Founded in 2020 and headquartered on Battery Street in San Francisco, Coalesce is, in plain English, the place where SQL pipelines get built without the SQL being copy-pasted ten thousand times. The platform is visual. It is metadata-driven. It writes a lot of code for you, on purpose. It now runs on Snowflake, Databricks, and Microsoft Fabric, which is to say - the three places enterprises actually keep their data in 2026.

There are 140 people working on it. They have raised about $84 million. Their customers include casinos, diners, oil majors, and investment banks. None of those facts, individually, tells you why Coalesce matters. Together, they start to.

The problem they saw

To understand Coalesce, you have to understand the unglamorous middle of the modern data stack. The fashionable parts are the warehouse (Snowflake, Databricks), the BI tool (Tableau, Looker, Power BI), and increasingly, whatever AI model is rendering a chart this quarter. The unfashionable part is the bit in between - the transformation layer, where raw data gets stitched into the clean, governed, version-controlled tables that everyone downstream actually uses.

That layer is mostly SQL. A lot of SQL. Written by humans. Reviewed by other humans. Then rewritten three months later because the schema changed, or because the business definition of "active customer" changed, or because someone got promoted. Anyone who has worked in data knows the genre.

Armon Petrossian and Satish Jayanthi knew the genre well. They met at WhereScape, a company that had been pioneering the slightly nerdy idea of data warehouse automation since the early 2000s. Between them they had worked, by their own count, with more than a thousand companies undertaking data migrations. They had seen what broke. It was usually the same thing.

The founders' bet

The bet was simple to describe and hard to execute. Take the parts of building a data pipeline that everyone does the same way - column lineage, change management, deployment, documentation - and let software do them. Take the parts that are genuinely creative - the business logic, the joins that nobody else would think of - and give engineers a beautiful place to write them.

It is a low-code platform that is, importantly, not no-code. The distinction matters. Coalesce is built for data engineers, not for executives who want to drag a pie chart onto a Zoom call. The graphical interface is a productivity layer over real, inspectable, version-controlled SQL.

The company spent two years more or less in stealth. They launched publicly in 2022 - which, in a market that had spent the prior decade on a vibe of "ship the demo, fix the rest later," counted as a long courtship.

The product, in plainer English

Coalesce now ships three things that matter. Coalesce Transform is the original platform - a visual, metadata-driven workspace where data engineers build pipelines using reusable templates, and the system generates the underlying SQL. Coalesce Catalog, added later, is the AI-powered metadata layer - a place to discover, document, and trust the data those pipelines produce. Coalesce Copilot is the AI assistant that drafts, explains, and documents transformations inline, because of course it does, it is 2026.

There is also a Marketplace, where users share packaged transformation patterns - the data engineering equivalent of an open-source npm registry, except the packages tend to be auditable governance templates rather than left-pad.

What customers actually do with it

The use cases sound dry until you say them out loud. Migrate twenty years of legacy Teradata logic into Snowflake without losing your mind. Stand up a governed data vault before the auditor visits. Build a metric layer that the finance team and the marketing team can both agree on - which is a corporate miracle. Document a pipeline so the analyst who inherits it next year can read it without crying. Coalesce sells these outcomes as time saved, but what it really sells is the absence of late-night Slack pages about a broken job.

The proof

The customer list is more interesting than the slogan. Caesars Entertainment uses Coalesce. So do CKE Restaurants and Denny's, which means somewhere in America there is a stack of pancake margin data flowing through it. So does Houlihan Lokey, the investment bank, and TotalEnergies, the French oil major. Casinos, diners, banks, and oil. A useful sample of the actual economy.

The investor list is also worth reading. Series B was led by Industry Ventures and Emergence Capital, with participation from Snowflake Ventures - the venture arm of, yes, the platform Coalesce was originally built on. That is an unusually direct vote of confidence from a partner. Bob Muglia, former Snowflake CEO, also wrote a check. There is a degree of insider conviction here that does not show up on a pitch deck.

The mission, stated plainly

Coalesce describes its mission as helping data teams build, document, and manage data transformations roughly ten times faster, without giving up governance and standards. That last clause is the one to underline. The pitch of the last decade in data was speed. The pitch of this decade is speed without the regulator finding out.

"Move fast and break things" never really worked in data engineering. Now that AI models are training and reasoning over enterprise data, it is no longer even cute. Coalesce's argument is that automation and governance are not opposites. They are, in fact, the same product if you build them together from the start.

Why this matters tomorrow

Every enterprise AI project has a dirty secret - the model is the easy part. The hard part is getting the underlying data into a shape the model can read, lineage you can audit, and tables your lawyers can defend. If transformation is broken, AI is broken. If governance is broken, the next regulator visit ends in a Bloomberg headline.

This is the bet Coalesce is making for the next five years - that the company best positioned to power enterprise AI is the one quietly fixing the unfashionable plumbing underneath it. Not the one selling the chatbot. The one selling the pipeline that feeds the chatbot a fact rather than a hallucination.

Back to that data engineer at the Fortune 500 retailer. The dashboard is correct. The pipeline runs at 3 a.m. and nobody is awake to watch. The audit log is complete. The model on top of it can tell the merchandiser, with reasonable confidence, what to put on the front page next week. There is no parade. There is just the next press of the button. That, more or less, is the product Coalesce is selling - and the very specific kind of revolution it represents.