Yochay
Ettun

He Started Debugging at Seven. Intel Came Calling at Thirty.

There is a version of the Yochay Ettun story that starts with the Intel acquisition headline and works backward. That version misses the point. The real story starts in a child's bedroom in Israel, with a kid who figured out how to make computers do things before he fully understood what the things were. That instinct - build first, explain later - never left him.

Ettun studied Computer Science at the Hebrew University of Jerusalem. While there, he didn't just attend lectures - he founded the HUJI Innovation Lab, a move that telegraphed exactly the kind of operator he'd become: someone who, when handed a platform, builds something on top of it. Before graduating, he was already CTO at Webbing Labs. Before most people his age had a second job title, he had three.

Four years in the Israeli Defence Force's Intelligence unit had something to do with the discipline. The IDF's tech corps is an unusual finishing school for founders - structured enough to instill rigor, demanding enough to filter for people who can solve hard problems under pressure. Ettun came out the other side not with a chip on his shoulder, but with a clearer read on what "hard" actually means.

cnvrg.io: Built for the Problem Nobody Had Named Yet

By 2017, Ettun and his co-founder Leah Forkosh Kolben had spent years doing AI consulting. They kept hitting the same wall: data scientists were brilliant at building models and terrible at shipping them. The infrastructure was a mess. The tooling was a patchwork of scripts, spreadsheets, and crossed fingers. More than 80% of ML models never reached production. Not because the science was wrong - because the plumbing was.

They'd built an internal tool to manage their own workflow. Then they looked at it and thought: every data science team on earth has this problem. So they turned the internal tool into cnvrg.io - a platform that handled the entire ML lifecycle from experimentation to deployment, on-premise or cloud, free tier or enterprise.

"Cnvrg is a platform built by data scientists for data scientists" wasn't a marketing line. It was a design constraint. Every feature had to pass a single test: would a working data scientist actually use this, or would they route around it? Customers like Lightricks, ST Unitas, and Playtika said yes. The platform raised $8 million from Hanaco Venture Capital and Jerusalem Venture Partners.

Intel said yes in November 2020. The acquisition - reported at around $60 million, though Intel never confirmed the figure - was framed as strategic: Intel had been doubling down on AI infrastructure, and cnvrg gave them a software layer that made their hardware more valuable. For Ettun, the appeal was scale. "We wanted to build something that is really big in AI. We wanted to make an impact in the AI world." A chipmaker's distribution does that faster than a Jerusalem startup's go-to-market motion ever could.

The Introvert Who Had to Learn to Sell

Ettun is quick to admit that the transition from engineer to CEO was not comfortable. "Doing sales demos and pitches to other potential customers was very hard for me," he said in an interview with CTech. That's not a boilerplate humble-brag. He means it literally: he is by nature someone who prefers to build rather than persuade, to debug rather than demo.

What he did - and this is the thing worth paying attention to - was treat the CEO role as a technical problem. Every quarter, new skills were required. So he learned them. Not fluently, not without friction, but methodically. "My job as a CEO is evolving every quarter. There are new skills I need to develop. It's fun and I love it." The framing is very much that of an engineer describing a project.

The Scout: Betting on the Next Generation

After the Intel chapter closed, Ettun surfaced at Andreessen Horowitz as a Scout - one of perhaps 21 scouts the firm has deployed across Europe and Israel as of 2025. The a16z scout program is not a training ground. It's a deployment mechanism: operators and founders who already see deal flow, given a check-writing mandate to act on it early. Scouts typically write checks of $10,000-$25,000 per deal. The point is access and pattern recognition, not capital deployment at scale.

For Ettun, the fit is obvious in retrospect. He spent years in the exact terrain a16z wants intelligence on: Israeli AI infrastructure, early-stage MLOps, the intersection of hardware and software that Intel also found valuable. In September 2024, he participated in Bluebricks' $4.5 million seed round alongside Flint Capital and Glilot Capital Partners - a bet on cloud infrastructure management, the kind of unsexy plumbing problem he knows intimately from the cnvrg.io years.

The LinkedIn update in late 2024 said it plainly: "It's great to be building again." Whether that means building companies, building a portfolio, or just building - the instinct from age seven is still running.

How the Story Unfolded

The Ledger

What He's Actually Said

Domain Knowledge

Further Reading