Preference Labs is an applied research and infrastructure lab for intelligent agents.

We build the operational layer agents need to reason and execute in live markets. Our systems combine domain ontology, structured tools, and execution primitives so models can move from raw signals to real decisions under production constraints.

Our research focus is pragmatic: what makes agent behavior coherent over long horizons, robust across venue changes, and adaptive under uncertainty. We treat market environments as high-signal testbeds where hypotheses can be stress-tested quickly and measured directly.

The result is infrastructure designed for external agent teams, not a single demo model. We are building systems that make live context legible, decision quality durable, and execution workflows reliable across evolving market conditions.