Physics-First Perception for Real Autonomy

FieldSpace was founded in 2025 by Jack Al-Kahwati, an aerospace engineer frustrated by the fundamental weaknesses of pattern-recognition approaches in autonomous driving. After years of observing industry reliance on black-box neural networks, Jack set out to prove that a physics-based approach—using only commodity cameras and first-principles mathematics—could deliver more reliable and explainable autonomy.

The initial concept came from direct experience with edge cases that consistently failed on conventional stacks. Jack started with a simple hypothesis: if humans can drive using only their eyes and an understanding of physical laws, then a machine should be able to do the same without expensive hardware or massive data training. He began prototyping in early 2025, building PDE-based algorithms that model how light and motion interact in real space, rather than relying on statistical guesswork.

Within months, a prototype was running live perception on standard hardware, using off-the-shelf cameras and custom PDE solvers. Early tests showed that this approach could handle previously unsolved perception failures—such as faded lane markings and variable lighting—with deterministic accuracy.

As FieldSpace took shape, Jack assembled a focused team of engineers and advisors, formalized the company, and started field trials with local partners. While the industry was still dominated by neural network approaches, FieldSpace demonstrated that first-principles physics could deliver explainable, robust autonomy.