Decades inside the tech industry

We were both at Redis as it scaled hardest — one of us architecting across teams, the other leading a core team under constant production pressure, tens of thousands of live databases. What stayed with us wasn’t the code. It was how teams hold together when the system can’t go down, and how you move people onto new practices when no one has time to stop. We saw the same company from two seats — part of why we trust each other’s read now.

At Ultima the hard part wasn’t the pipelines — it was that backend, frontend, ML, and data-science engineers were effectively speaking different languages and drifting into silos. I spent as much time building a shared vocabulary as writing code, so people could collaborate instead of talking past each other. Most of what we do now is a version of that: getting groups who each think they’re right to understand each other.

At Elbit I spent years bringing object-oriented methods into real-time defense products — into an engineering culture with deep, established habits. The technical case was the easy part. The real work was helping experienced engineers adopt a way of working they hadn’t chosen, without making them feel overruled. That’s the same challenge behind most “resistance” I’m called in for today.

Sunbit needed a real-time platform to capture every purchase as it happened, and brought me in to architect it — which meant making the case for technologies the team hadn’t used, like Kotlin and Kafka Streams. New technology is rarely the hard part; getting a team to trust it is. I proved it in a proof of concept, then wrote production code alongside them until it was theirs, not mine.

Line5 builds physical AI — robots running computer vision in the field. I joined the early team as an architect, from the GPU vision pipeline to the call on whether to build on ROS, or a lighter layer like Zenoh. The harder problem was human: backend, frontend, and ML engineers, each with their own standards, barely agreeing on what good code looked like. So I built shared ground — common libraries, shared data types, and coding standards that held for the AI coding agents I brought onto the team as much as for the people. That’s what keeps a field robot debuggable when something breaks.

Band builds infrastructure for AI agents to talk to each other; I came in as a software architect, consulting on specific problems — prototyping with advanced AI coding agents, including how to scale their database across many nodes. But the CTO also pulled me into something harder: an engineer who wasn’t delivering, no manager between them, a breakdown where each felt unheard. I gave them a straight read — skills gap or communication gap — then helped them hire the manager the team had been missing. Being credible in their code is what made the read on their people land.

Enigma was an early-stage startup with no real development process and a team that needed to grow fast. I built the process from scratch and, harder, built a culture where people felt safe to lead and to disagree openly. Seeing how fragile and how decisive that culture was taught me how much of a team’s output comes down to whether people trust each other — which is where I start now.