Somewhere in an engineering organization, a perfectly intelligent person is hunting through a perfectly static PDF. The drawing is modern. The product model is modern. The procurement system is modern enough to have opinions. Yet a requirement governing a material, test or process is still sitting inside prose, waiting to be found, interpreted and copied by hand. Rupert Hopkins made a company out of this peculiar mismatch.
The company is XSB, founded in Setauket, New York, in 1998. Its territory sounds dry until the consequences arrive: obsolete parts, missed changes, duplicated research, questionable prices and specifications that point to other specifications like a bureaucratic treasure map. Hopkins's central observation was that the industrial world had digitized documents without truly digitizing the knowledge inside them. A PDF can travel instantly and still behave like a sheet of paper.
His answer was to stop treating the whole file as the useful unit. A requirement, a table, a material and a reference could each become data with an identity. Each concept could have its own address. Connections between those concepts could then be queried, tracked and reused by people and software. XSB called its platform SWISS, short for the Semantic Web for Interoperable Specifications and Standards. It was a deliberately grand name for a very practical ambition: let the words that govern engineered things participate in the digital thread.
“We call these smart, connected documents. You can think of this as the Internet of Things for the concepts inside a document.”Rupert Hopkins, describing SWISS in 2017
The long route to one stubborn problem
A career built around useful information
Hopkins did not begin with a software founder's standard origin story. From 1987 to 1995, he served as Long Island Regional Director of Economic Development for New York State. The work placed him close to manufacturers, employers and the public machinery designed to help a region grow. In 2011, years after leaving that post, he returned to the regional economic-development conversation as a member of Long Island's council, listed alongside leaders from Stony Brook University, Brookhaven National Laboratory, Northrop Grumman, Verizon and local businesses.
In 1995 he joined Dayton T. Brown to direct the Center for Alternate Sources of Parts, known as CASP. The center was backed by DARPA and addressed a severe post-Cold War logistics problem: what happens when a needed component disappears from the market, or its original producer no longer makes it? CASP connected a network of specialized manufacturers with a sourcing system that used the XSB Prolog logic engine to extract and infer information about parts and the companies capable of making them.
That experience supplied the bridge to XSB. Alternate-part sourcing is not simply a search exercise. A replacement must satisfy technical requirements, manufacturing constraints and sometimes regulatory rules. The puzzle is made of relationships. Hopkins left Dayton T. Brown in 1998 to found a company equipped to model them.
Begins eight years as Long Island Regional Director of Economic Development.
Moves into defense logistics as director of the Center for Alternate Sources of Parts.
Founds XSB in Setauket with a focus that will expand from parts data to connected engineering knowledge.
Exiger acquires XSB after 26 years of independent company building.
Documents, dismantled
The useful trick is changing the unit
The technical lineage mattered. XSB was founded by data scientists connected to Stony Brook University, and its work drew on the XSB Prolog language developed by co-founder David S. Warren and his academic collaborators. Logic programming, ontologies and semantic links became less a laboratory curiosity than a way to answer an engineer's ordinary questions with less rummaging.
Consider a technical standard. To a reader, it is a sequence of pages. To the model Hopkins and his colleagues patented, it can become a structured collection of governing authorities, subjects and requirements. The document is converted into a syntactic model. An ontology-guided process identifies meaning. The result can be queried by a designer choosing a part, a maintenance engineer planning a repair or a purchasing manager placing an order. Different people can receive the slice they need without retyping the same knowledge into another system.
The unglamorous detail is the product. A reference can remain connected to its source. A tailored work instruction can be assembled from authoritative passages. When a source changes, the system can identify the downstream material affected by that change. The promise is not a robot engineer. It is fewer hours spent discovering that somebody copied an old paragraph into a live process.
Hopkins explained the day-to-day benefit in an interview about engineering productivity: a product specialist should not have to visit the engineering library, download a specification, read it, interpret it and then repeat the journey for the next test. If the information is interoperable with the engineer's existing systems, it can appear at the moment of need. XSB's own operating principle put the same thought more bluntly: engineers and procurement professionals do not need another workbench. They need answers inside the workflow they already use.
“SWISS offers a big boost in productivity for product specialists working on any particular design.”Rupert Hopkins on connected engineering documentation
Scale without spectacle
AI as an enabler, with paperwork under its fingernails
Long before generative AI put a chat box on every product roadmap, XSB was using machine learning and semantic technology to classify parts, extract attributes and link concepts. The company's language around that work is notably sober: “AI as Enabler, Not as Magician.” For a business serving engineering, procurement and government users, the distinction is commercial common sense. A charming guess is useless when the question concerns a regulated material, a counterfeit component or whether an obsolete part has a qualified alternative.
Context is the valuable ingredient. XSB combines proprietary, public and commercial information, cleans and standardizes it, then delivers the result into enterprise applications through APIs. The task is not merely to retrieve a plausible paragraph. It is to preserve the relationships that explain which authority governs which subject, and which requirement must be satisfied. The cleverness is hidden so that the answer can be used.
By August 2024, when Exiger announced its acquisition of XSB, that quiet infrastructure had substantial reach. The catalog covered more than 100 million parts and 400 million attributes used by the federal government, from camouflage clothing to weapons systems. Thousands of engineers, item managers and catalogers used the product suite for problems including obsolescence, nonconforming parts, pricing, availability and regulatory compliance.
For Hopkins, the combination was about range. XSB had sourcing, logistics and sustainment tools. Exiger brought broader supply-chain mapping, orchestration and risk analysis. In his announcement, Hopkins said XSB's AI and knowledge base would “flourish” inside the larger software ecosystem. The deal offered an institutional home for work that had spent more than two decades moving between defense logistics, engineering systems and standards publishing.
Patience, patented
The market eventually learned the vocabulary
The patent trail tells its own story about the pace of infrastructure. A filing with priority dating to December 2012 described systems for creating, editing, storing and retrieving knowledge contained in specification documents. Hopkins was named alongside David Winchell, Louis Pokorny, David Warren, Tatyana Vidrevich, Christopher Rued and Raymond Weidner. The first U.S. grant arrived in June 2020. A continuation was granted in January 2025.
Between those dates, Hopkins kept carrying the idea into rooms where standards and engineering practice meet. He presented on semantic value at a 2017 event for standards publishers. That year he authored a conference paper explaining how every concept in a document could receive a unique address. In 2018, he and Vidrevich described linking technical requirements beyond the product-lifecycle-management vault. In 2022, Hopkins, XSB board member Bob Solomon and Andrew Bank won Best in Show in the World Standards Day paper competition for a proposal on the digital transformation of engineering standards.
Their recurring argument was simple enough to survive the acronyms. Industry had transformed two-dimensional drawings into rich digital models, but the documents governing parts, materials and processes still behaved much as they did in the 1990s. A digital twin document would treat text, tables, equations, images and requirements as interoperable data points. Machines could read more of the burden; people could spend less time reconciling it.
Hopkins's career is best understood as a long campaign against information that is technically available and practically inert. Economic-development offices hold knowledge about companies. Logistics programs hold knowledge about suppliers and parts. Standards contain knowledge about how things must be designed, built and tested. The recurring opportunity is to make those facts useful at the moment a decision is made.
There is a founder's lesson here, tucked inside the paperwork. Important markets do not always announce themselves with novelty. Sometimes they are visible in the manual step that everyone accepts, the old file format nobody loves, or the requirement copied for the ninth time because the systems on either side cannot speak. Hopkins chose one of those seams and stayed with it while the language changed from expert systems to the semantic web, from the Internet of Things to knowledge graphs and AI.
The nouns evolved. The expensive problem remained. Somewhere, an engineer still needs the right requirement for the right part before the wrong version finds its way into a real object. Hopkins's work asks the document to do more than sit there looking official. It should know what it contains, where those ideas came from and what else will move when one sentence changes. After all, even the driest specification has a social life. XSB simply taught the machines to follow it.