Hans Salomonsson has lived quant trading in London to NASA Goddard to Chalmers, where he met Prof. Devdatt Dubhashi - Cornell PhD, Max Planck postdoc, OECD AI advisor. They spun Embedl out of an EU research project in 2018, on the bet that the AI revolution won't be won in the cloud but on devices.

A self-driving truck generates terabytes per hour and makes life-or-death calls in milliseconds - an eternity at 70mph. The AI must live on the vehicle. But edge hardware is brutally constrained: tiny batteries, no cooling, limited memory - and cloud-grade models are 10-100× too big.

Embedl compresses AI to fit on anything - they've shown 83% energy savings, 95% memory reduction, 18× faster inference. Hardware-agnostic. Kodiak Robotics (now public) runs Embedl on its driverless trucks in Texas, just across the pond. So do Bosch, Volvo's Zenseact, and Tobii.

In 2024, NVIDIA paid $300M for Deci AI and $165M for OctoAI - both in AI model optimization, but for cloud. Embedl does it for the edge, inclusive of drones and defense. They just raised last June and racked up some awards along the way - and with only $12M in total funding, they're now raising their A.