System design for physical AI: a patch cycle that learns from every confirmed call
By Muhammad Umar. This is the engineering summary of an open reference architecture. The paper, its object model as JSON and the model register are free to reuse under CC BY 4.0: https://codeatoms.ai/ot-vulnerability-utility-us/ . DOI to follow on Zenodo. The operator is described by class, never by name. An electric utility in the United States runs substations, control centers and generating…
The paper outlines a system design for physical AI in critical infrastructure protection, specifically for electric utilities in the United States. The system relies on a read-only model of assets and only allows outbound data under specific conditions. Seven systems of record feed into the design, providing information about assets, vendor advisories, national vulnerability advisories, change management, work orders, evidence repositories, and network monitoring tools.
All records carry their source system and freshness information, with business-critical drafts refusing data older than a certain age. Apache Kafka 4.4 acts as the backbone, relaying information between adapters and the ontology, while idempotent upserts ensure data safety and prevent duplicate decisions. The system consists of sixteen objects with thirteen typed links, with three objects created within the design: applicability decision, disposition, and cycle agent.
The inference model, a single node with eight high-end GPUs, processes the data and provides the necessary outputs, such as applicability decisions, disposition, and evidence items. The entire system is built on open-source, permissive licenses and relies on domestic, non-export controlled hardware for US-based users.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
