Yochai's conviction was formed at a control station, in the middle of an emergency. As a mission-critical operator, he experienced a malfunction that nearly ended in disaster. Every system around him could display telemetry and retrieve procedures, but none of them could truly own the situation.
Having spent more than a decade operating high-stakes systems before building mission-critical command-and-control software, Yochai had lived on both sides of the divide: as an engineer and as an operator. The gap was clear. The world's most advanced software could show people the problem, but it could not reason about it.
Then the AI wave arrived, making that gap impossible to ignore. Frontier models could generate images and perform increasingly sophisticated digital tasks, while the control rooms running power plants, data centers, and aircraft still relied on fragmented intelligence. Sensors showed what was happening, while documents, diagrams, charts, and veteran operators held the knowledge of why it was happening and what to do next.
Those veterans are retiring faster than they can be replaced. Together with his brother, Matan, who spent nearly seven years at Google working across large-scale infrastructure and AI research, Yochai founded Psistar around a clear thesis: AI for mission-critical environments must be physics-native. It must, as they say, "speak physics."
That means building AI around a real understanding of physical dynamics, grounded in verified operational knowledge and deterministic decision paths, capable of operating on-premises without requiring new sensors.
Over time, the scale of the challenge became even clearer. AI's growth is gated by energy, energy is gated by infrastructure, and every new facility creates more surface area for downtime and more opportunities for optimization. As the world builds thousands of new power plants and data centers, someone has to operate them efficiently.
Psistar's answer is to enable one-to-many operations, turning scarce operational expertise into an industrial reasoning system that understands a facility's physics, forecasts its future state, and acts safely within approved procedures.
The next decade belongs to the real, noisy, physical world. Yochai's ambition is to help run it well.