A data warehouse is an odd place to look for impatience. Its name suggests shelving, orderly aisles, things put away for later. Eldad Farkash has spent much of his career worrying about what happens when later arrives. Someone asks a question. The data exists. The answer takes its time. Between those two events sits an engineering problem that has followed him from Sisense to Firebolt, changing shape as the people asking the questions change.
His work has moved between business intelligence, cloud infrastructure, company leadership, and the microphone of a technical podcast. The recurring subject is the same piece of friction: how much effort stands between a person and something useful in their data? That makes his career more coherent than a list of job titles would suggest. The titles explain where he sat. The queries explain why he kept coming back.
A business user at the other end
Farkash was born in Ra’anana and grew up in Savyon. As a teenager, he wrote code on demand with a childhood friend, Ofer Israeli. Their fledgling operation attracted a lawyer’s letter over its name before they had formally created a company. There is a certain entrepreneurial efficiency in acquiring a legal dispute before acquiring brochures.
He completed bachelor’s and master’s degrees in computer science at the Interdisciplinary Center Herzliya. A room full of servers in his father’s business helped sharpen the problem that would lead to Sisense. Expensive equipment was being devoted to getting answers from company data. The person depending on those answers mattered as much as the equipment producing them.
By December 2014, Farkash was describing Sisense through the needs of business users. The company joined data preparation and dashboards in one stack, so customers could assemble an answer without stitching together a procession of vendors. The ambition was practical: let a person with a business problem work through it without turning every question into a lengthy infrastructure project.
One example involved eBay’s shipping analytics. Information was distributed across continents, and the business needed a unified view of products, vendors, seasons, and geographies. In Farkash’s account, people responsible for business analytics could combine the data themselves. The interesting outcome was who could do the work. A tool changed the boundary between waiting for another team and pursuing your own question.
An award for looking inside the machine
The polished dashboard depended on decidedly less decorative work. Farkash helped develop Sisense’s In-Chip analytics, an approach built around using processor capabilities to handle analytical work efficiently on commodity hardware. In 2013, that work earned him the World Technology Award in the individual IT Software category.
At the time, he described an ambition to make analysis of large datasets available to businesses with a range of budgets. It was a concern about access expressed through system design. When infrastructure demands become smaller, more organizations can attempt the work. The technical decision reaches all the way into the customer’s purchasing decision.
“It was a real honor to win this award.”
Eldad Farkash, on the In-Chip award
His examples ranged well beyond office reports. Galaxy Semiconductor used analytics to visualize wafer quality, turning billions of records into heat maps. A simple picture could require thousands of SQL queries. That is the peculiar bargain of an effective interface: the clearer the result looks, the easier it is to overlook the amount of work required to put it there.
Farkash’s career sits inside that bargain. The visible object might be a colored wafer, a shipping dashboard, or an answer on a screen. Beneath it lie decisions about how data is arranged and how computation is performed. Making those decisions well changes what the person at the screen can reasonably attempt next.
Speed boats in a sea of battleships
In 2014, he was already advising startups to look for a different shape of solution rather than compete with large infrastructure companies on their own terms. His nautical suggestion was compact: “focus on speed boats rather than battle ships.” The space between those words carries much of his engineering outlook.
He questioned whether problems that appeared to require vast clusters could be reduced to something far smaller. Hardware improvements offered openings. So did software that made complicated infrastructure easier to manage. His concern was the burden placed on the organization trying to operate the system, as well as the headline amount of data it could hold.
That same period brought a longer view of startup work. He urged founders to build for at least five years of changing technology. Quick assembly could produce a working solution while leaving its makers with a difficult technical future. The advice asks an uncomfortable question of a young company: will the design remain useful after the trend that made it fashionable has moved on?
He also watched new data sources emerge from machines and the surrounding environment. Open-source projects, universities, and large technology companies could keep reopening a supposedly mature market. In that view, experience needs to remain alert. Knowing how the previous generation worked helps only if it leaves room to notice what the next one requires.
Back down to the engine
Firebolt, founded in 2019, gave Farkash another place to pursue the problem. He co-founded it with Saar Bitner and Ariel Yaroshevich, fellow Sisense alumni. His explanation for starting again concentrated on the layer that actually computes a query. After years of building an entire business-intelligence experience, he wanted to spend his attention closer to the engine.
In his 2020 account of the new company, storing data had become less of an obstacle than turning it into timely answers. He also questioned the incentives of cloud billing: if customers pay for resources consumed, what encourages a provider to help them consume fewer? Efficiency, in his telling, had to make commercial sense as well as technical sense.
From the business dashboard to the application itself, the useful measure is how long a person waits for a query to return.
His 2021 description of Firebolt focused on indexes that identify relevant ranges of data, with computation scaling separately from storage. A warehouse could retain one version of the data while different computing resources served different needs. The attraction was doing the needed work efficiently, rather than treating every request as a reason to move and inspect everything.
The audience was changing, too. A website, game, or application might use analytics as part of the experience it delivers. The person asking for data could be a product user rather than an analyst preparing tomorrow’s report. That puts query processing on a different clock. A delayed internal answer is inconvenient; a delayed interaction becomes part of the product.
The brothers with the follow-up questions
There is another way to encounter Farkash’s interest in all this: listen to him ask someone else about their work. The first season of The Data Engineering Show paired him with his brother Boaz. Later, Benjamin Wagner joined him. The show’s premise is practical conversation with people responsible for making data systems function.
In an episode with Explorium’s Roy Miara, the discussion moved through multiple data sources, evolving schemas, engineering responsibilities, and the tools underneath the product. The opening also contained a simple exchange between the brothers: “I missed you.” “I missed you so much.” It is a pleasant introduction to a conversation that will eventually concern database architecture.
Farkash’s questions then went straight to how the system was organized. Was the data one source? One schema? How did users add outside information to their models? The details make the exchange useful. A general assertion that a company works with data leaves most of the engineering story untold. Asking how the pieces meet begins to uncover it.
With Wagner, he also hosted Statsig founder Vijaye Raji for a conversation about observability and its business consequences. The questions reached into how companies evaluate features and learn from user behavior. That is a natural extension of the earlier work: once the system can deliver an answer, people still need to choose a worthwhile question.
Capital, colleagues, and another chapter
Firebolt’s January 2022 financing put a dollar figure on investor expectations: a $100 million Series C at a $1.4 billion valuation. The same announcement brought another consequential name, Mosha Pasumansky, a former Google BigQuery principal engineer who became CTO. For a company selling an analytical engine, the people capable of building it belong in the story alongside the capital.
Farkash’s connections also extend into investing. Angular Ventures identified him among the investors joining its backing of FalkorDB in 2024. The company works with graph databases, another part of the larger effort to make information usable. For Farkash, it adds another database company to a working life already spent with several kinds of analytical system.
In September 2025, Firebolt announced that Hemanth Vedagarbha would succeed Farkash as CEO and that Farkash would become chairman. He welcomed the expansion of Vedagarbha’s leadership after eight months running go-to-market. The shift placed another executive in charge of operations while giving the co-founder a different role in the company’s next phase.
The job title changes. The question remains.
The company’s subsequent path has included difficult decisions. In February 2026, Firebolt cut dozens of jobs in Israel and abroad during a broader shift toward efficiency. Its current leadership page lists Wagner as CEO and Farkash as co-founder. These changes belong beside the funding milestone. A valuation records a moment; running a company keeps creating new ones.
Firebolt now describes an analytical database for engineers and agents, with open standards and deployment choices extending from a laptop to cloud clusters. The language has evolved from the earlier warehouse pitch. The concern with fast answers and manageable infrastructure remains recognizable.
Farkash’s public career gives that concern a human outline: the business user assembling a dashboard, the engineer deciding how a query runs, the podcast guest explaining what broke, the founder handing over a job. Each occupies a different position around the same machinery. The useful question, after all those years, is still wonderfully ordinary: how soon can I have the answer?