A perfectly good number can cause an astonishing amount of trouble. Put it in a dashboard, call it revenue, and send it to three departments. The arithmetic may be impeccable. The argument begins with what everyone thinks the word means. Sarah Levy has made a company out of the work that follows: finding the definitions, decisions and habits behind the figure.
Levy is the co-founder and CEO of Euno, which she started with Eyal Firstenberg in 2023. The company began by addressing the strain between analysts’ freedom to build and data teams’ responsibility to keep business logic consistent. Its work now extends to the organizational context AI agents need. Along the way, Levy has kept returning to a practical question: how does a growing organization preserve shared meaning while its people keep changing things?
It is a question with very little glamour at first glance. Nobody has ever asked for a commemorative photograph of a well-governed metric. Yet the consequences reach straight into the decisions companies make. Give a system access to information, and you have solved the access problem. Getting it to understand the company remains an entirely separate piece of work.
A physicist’s appetite for inconvenient questions
Levy’s route here passed through physics and military technology. She joined Talpiot, the Israeli military’s technology training program, at 18. The program combined a degree in physics and mathematics at the Hebrew University with military training. She graduated at the top of her class and went on to lead technology projects and a cybersecurity department during more than a decade of service.
She then chose to study theoretical physics at the Weizmann Institute of Science, earning her master’s degree in 2013. She had already taken graduate-level courses while serving. Concentrated study, she decided, would require its own time. That decision is a useful detail in a career otherwise easy to compress into a procession of leadership jobs: she made room to be a student again.
Her subsequent work included senior research and development leadership and a CTO role at Sight Diagnostics, followed by data science and analytics leadership in Pagaya’s real-estate business. The jobs placed her inside different organizations with different demands. Data crossed those boundaries. So did the difficulty of turning it into a dependable basis for action.
The office dictionary nobody quite shares
Consider an active user. A person who signed in? Someone who completed a transaction? A customer who returned within a particular period? Each definition may serve a sensible purpose. Trouble arrives when several purposes acquire one label, and the resulting figures circulate as though they describe the same thing.
This is the territory Levy and Firstenberg chose for Euno. Both had dealt with homegrown approaches to managing definitions and calculations. They wanted to help organizations keep business logic consistent without requiring analysts to stop experimenting or abandon the tools they used. A company can scarcely ask its analysts to discover something new while insisting that every possible discovery has already been centrally approved.
Euno’s early approach connected that work across business intelligence tools and the data layer, including dbt, where transformations can be defined in code. In her public launch message, Levy described integrations with Looker and Tableau and automation to detect, map and synchronize changes. The ambition was to make collaboration part of the workflow, with useful new logic finding its way back into a governed model.
In her conversation with Richie Cotton, Levy described the importance of understanding how data models are actually used. A business can accumulate calculations and tables faster than it can evaluate them. Keeping every experiment forever creates its own administrative comedy: the office cupboard eventually becomes the office. Usage helps distinguish a passing exploration from something the organization should maintain.
A post, a message, a former Tableau CEO
One of Euno’s advisory relationships began with Levy writing about these problems in public. Mark Nelson, the former president and CEO of Tableau, read her LinkedIn posts about balancing self-service analytics and governance. He contacted her. After hearing the pitch and seeing a demonstration, he joined the advisory board.
Nelson recognized a difficulty he had repeatedly encountered with customers: the data team owns centrally managed resources, while business users often hold the context that makes those resources useful. Asking either group to solve the entire problem alone leaves something missing. The connection gave Euno an adviser who knew the problem from the other side of a widely used analytics product.
It also gives a more concrete account of networking than the usual instructions to attend things and collect names. Levy had articulated a position. Someone with relevant experience recognized it and wanted a conversation. Public writing became a way to locate a collaborator.
Hiring on a longer clock
Levy has described a similarly deliberate beginning to the company itself. Before hiring their first employee, she and Firstenberg spent six months interviewing data leaders from more than 300 companies. Their own experience supplied the starting point; the conversations tested how far it extended. Familiarity with a problem can be helpful. It can also make the first explanation rather too comfortable.
That patience appears in her account of recruiting. She pursued one R&D leadership candidate for more than a year. Her reasoning was that the relationship would last much longer than the immediate inconvenience of waiting. A hiring process can be measured in weeks while its consequences are measured in years.
“You’re not hiring for six months; you’re hiring for years.”
Sarah Levy, on building a team
In discussing leadership with Lindsay Murphy, Levy brought family life into the conversation as well as company-building. A mother of three, she argues for workplaces that account for people’s responsibilities outside work. Her examples include extending flexibility to fathers and questioning whether late-night social events suit everyone. Culture has to survive an ordinary calendar, including the parts that never make it into a recruiting presentation.
Her advice to candidates is practical: talk to people already working at the company and assess the fit in both directions. The prospect of a job can make approval seem like the only objective. Levy asks people to consider the colleagues, management and everyday environment they are choosing, too.
The colleague who never asks around
By December 2025, Levy was describing Euno to Snowflake as a metadata intelligence platform for governance automation and AI context. Her focus was the gap between a promising analytics pilot and a system the organization could rely on. Existing metadata, she argued, could help agents identify useful, trusted data without imposing another large manual documentation project on the team.
The distinction matters because the next consumer of a company’s data may be software acting on someone’s behalf. An experienced colleague knows whom to ask when a figure looks odd. An agent needs the relevant knowledge supplied in a form it can use. The informal route through the office has to acquire a technical equivalent.
In September 2026, Levy described how human intervention had helped older data systems function despite gaps in documentation and lineage. Someone recognized a strange result. Someone remembered a changed definition. Moving to faster automated consumption changes the consequences of leaving those judgments implicit. Her argument is that the context around data needs to keep pace with the systems using it.
There is a human continuity here. Euno’s early work concerned collaboration between analysts and engineers. The newer work concerns the knowledge an agent needs to participate reliably. Both turn on the difference between possessing a table and understanding the decisions embedded in it.
The dashboard still needs an answer
On September 9, 2026, Euno announced a $23 million Series A led by N47, with participation from 10D and technology founders. The announcement put total funding at $29 million and described a team of nearly 30 across the United States and Israel. Levy’s accompanying argument centered on institutional knowledge: access to similar models would leave companies competing partly on their own accumulated understanding of how work gets done.
Led by N47. Euno reported $29 million in total funding.
Later that month, she presented at dbt Summit with Brad Levy, AlphaSense’s director of analytics. Her follow-up described a customer working on AI adoption alongside changes to its data architecture. The point was to make progress while improving the environment, rather than wait for a comprehensive reorganization to be finished. Anyone who has watched a company reorganize will appreciate the scheduling problem.
Her October 14 essay returned to the dashboard. Making a query visible, she argued, helps people inspect an answer, but it does not establish whether the query uses the right business definition or an appropriate table. Readability is useful. Reliability still requires knowledge of the environment behind the query.
This is where Levy’s story keeps landing: on the work surrounding an apparently simple answer. Her career has taken her through scientific training, technology leadership and the founding of a company. The recurring interest is in difficult problems that resist a convenient boundary. The number arrives neatly in its box. Levy wants to know how it got there, what it means and who should trust it.