Who you would be working with.

Twenty years of making an organization's rules machine-readable

Ulrich Callmeier is a computer scientist. He chose Saarland University for its focus on artificial intelligence and computational linguistics: language was the most interesting thing he could imagine understanding with computers, and AI, even with the methods of the 1990s, the most promising way to try. He worked in language-technology research at DFKI in Saarbrücken from 1998, and his diploma thesis produced PET, a parsing platform for large grammars that became one of the standard parsers of the DELPH-IN research community. He published on it in Natural Language Engineering (2000), in the CSLI volume Collaborative Language Engineering (2002) and, with Stephan Oepen, at IWPT 2000.

In 2000 he went to Silicon Valley, as a language engineer at a start-up building an AI system that answered customer-support e-mail. The system had to know the organization it was answering for, which makes it an early encounter with the problem Electric Mind works on today. PET went with him and became a substantial part of the company's processing stack. After two and a half years he returned to Saarbrücken for a short stint on research projects, among them SKATE, the SAP-funded language-checking research that Acrolinx later turned into a product.

In July 2003, while Acrolinx was still being spun out of DFKI in Berlin, he joined as CTO, before its first product release. Acrolinx makes AI content governance: quality gates, embedded in more than thirty authoring environments, that check enterprise content against each organization's own terminology, standards and rules. He stayed twenty years, until the end of 2023: CTO from 2003, managing director from 2009, and from 2015 also COO, which put finance, IT, support and customer success on his desk alongside product and engineering.

The job, for most of those years, was taking what an organization knew about its own language and making a machine apply it inside the tools people used every day, and then earning those people's trust in the result. That is an earlier version of what Electric Mind does with generative AI, where the context is no longer just language rules.

He founded Electric Mind in January 2024 as the home for everything he wanted to do next in AI: advising organizations on their initiatives, and working alongside a few founders on theirs. The two feed each other. What those projects teach about making AI hold up in real use goes into the consulting, and the other way round.