Modern biotechnology is organized by modality. Small molecules, antibodies, peptides, cell therapies, gene therapies, vaccines: each is a category with its own discovery methods, manufacturing stack, and regulatory path. We treat these categories as if they describe the natural structure of medicine. They do not. They describe the order in which we learned to make things.

A modality is a bundle of engineering constraints that happened to be solvable at a particular moment. Antibodies became a modality because we could express and select them long before we could design proteins from first principles. mRNA became a modality once we could manufacture and deliver it at scale.

We think the next era of biology will be modular rather than modal. A single protein is the most astonishingly versatile piece of machinery ever devised; they can sense, switch, transport, bind, and catalyze, all built from the same basic building blocks. The opportunity is to treat biomolecules the way engineers treat mechanical and electronic parts, and to build with them at the level of assemblies rather than single molecules i.e. nanomachines.

When many parts are composed in a modular, functionally decomposed way, the system acquires capabilities (emergent properties) that no single molecule has, the way a robot is more than its actuators and sensors. Scaling these assemblies, in the number of distinct parts, in physical size, and in the sophistication of their coordination, is the central technical problem we work on.

Getting there is a design problem and a measurement problem at once. On the design side, we build and scale AI systems that reason about parts and how they compose, so that we can specify an assembly the way an engineer specifies a circuit and predict how it will behave before it is ever made. On the measurement side, we build the experimental infrastructure to construct these assemblies and return what we learn to the models. Neither half works without the other. Models without tight experimental loops drift away from reality; experiments without models do not compound.

This is a long project and there are many unknowns ahead. If it is the kind of problem you want to spend years on, we are hiring.