Lattice Protocol
AI-native infrastructure for agentic workflows — federated compute, verifiable science, and composable modules that bridge human expertise with autonomous AI.
Lattice Protocol builds the plumbing for a world where AI agents and human scientists collaborate as equals. Modules compose into lattices, lattices federate across nodes, and every result carries a verifiable provenance chain.
The architecture grew from a realization at Google: cybernetic systems — human plus machine, data-driven, optimizable — can coordinate massive populations toward shared goals. The same patterns that coordinated fifty thousand translators could coordinate researchers, patients, and clinicians across institutional boundaries.
Three composable primitives make up the protocol: Modules (atomic capability units — one function, one model, one tool), Datasets (data with lineage, multi-cloud storage, federation metadata), and Lattices (directed graphs connecting modules into executable workflows). Lattices compose, nest, and federate across nodes. The goal is straightforward: compute should never be rare for rare disease researchers.
The project is open source. The mission is to make federated research infrastructure available to anyone building tools for patient care, diagnosis, and medicine discovery — so that the next patient navigating a diagnostic odyssey has tools built by someone who understands what that odyssey costs.