Leadership.

Rich spent his PhD at Duke on one question: how do you know a language model's judgment is right when the experts who label the data disagree? That work, on evaluating LLMs and on using them as judges, was published at venues including ACL and CHI, and he was the founding student of Duke's NLP lab. A lease is the same problem. Two competent attorneys will disagree about what a term is worth, and the model has to be measured against both.

Before Acrebase he built LLM infrastructure at Grammarly and Superhuman, where a model's suggestions reach millions of people every day. He serves on the Duke University Board of Trustees.

Fred has spent his career protecting systems that attackers actually go after. He was a staff cybersecurity engineer at Comcast and NBCUniversal and a security architect at Accenture, where he went on to lead generative AI development. He came to AI from security rather than the other way around, and Acrebase is built in that order: encryption and access control first, then the model.

A three-time founder, he runs engineering and infrastructure and does not outsource the part he knows best. The penetration testing is done in-house, and every action the AI takes lands in the audit log.

Andy has represented Texas developers for twenty-five years, on the landlord side of the shopping-center leases Acrebase models. A shareholder at Nathan Sommers Gibson Dillon in Houston and a member of the firm's management committee, he advises developers on acquisition and leasing and regional lenders on their credit facilities. He began his career as a corporate securities attorney at Cahill Gordon & Reindel in New York.

The deal-point taxonomy is his: which terms matter in a Texas retail lease, where each side starts, and what actually gets signed. He reviews what the model produces the way he would review an associate's redline, and nothing reaches a customer before it has passed that review.