I build reinforcement learning systems for real-world logistics operations.

Container terminals are among the most complex operational environments on earth.

Every day, thousands of containers move through a terminal. Each one has to be placed, stacked, and retrieved in a sequence that accounts for vessel stability, departure schedules, crane availability, yard capacity, and dozens of other constraints. A single bad stacking decision cascades into hours of wasted crane moves and delayed departures.

The industry has relied on heuristics and manual planning for decades. These methods break down as terminals scale and schedules tighten. With billions of possible configurations for a single vessel, the combinatorial complexity makes it a natural fit for reinforcement learning, but bridging the gap between RL research and actual terminal operations is where most approaches fail.

From research to real-world systems.

Through loadmaster.ai, I've taken reinforcement learning from the lab to real terminal environments. The systems I build don't just optimise in simulation. They're being deployed at actual container terminals.

Deep research foundations. Deployment experience. Domain understanding.

Most ML teams that tackle logistics treat it as a generic optimisation problem. They lack domain understanding. Most domain experts in terminal operations don't have the ML depth to build what's needed. I sit at the intersection: a research background in deep RL and signal processing, combined with years of working directly with terminal operators to understand the constraints that textbooks don't cover.

The path here wasn't conventional: from the British Army to plasma physics research at Princeton and Max Planck, to a PhD in AI-driven medical imaging at TU Eindhoven, to building RL systems for one of the world's most complex industries. Each step built a different kind of understanding. Together, they're why the systems I build actually work in the real world.