Deterministic execution systems, sovereignty inversion in LLMs, protocol-first AI, constrained inference, and reproducible ML pipelines.