Credasis, a North Dakota company building tools that help people trust data when it does not come with labels, is advancing prototypes across industrial control, network security, and quantum verification, posting competitive results on public technical leaderboards, and developing a new AI encoder centered on security and fidelity. Co-founders also joined the Fargo founder community on stage this week.

On the Fusion Equilibrium Challenge (NeurIPS 2026) Phase 1 development board, Credasis (verity_cpu) ranks #12 on Challenge 1 (DIII-D), and is the only entry in the top 20 with the Compute Light badge (as of Sept. 12, 2026; rankings change).

On the public EvalAI FALCON H1 leaderboard, Credasis AI (VERITY fsu_cpu) ranks #10 overall and #2 in Few-Shot Unsupervised, and is the only CPU-labeled entry. It is also the only team whose held-out score improves on held-in (as of Sept. 12, 2026; rankings change).

Credasis co-founders Tony Mangnall (Co-Founder and CEO) and Marty Oelrich (Co-Founder and Chief Scientist) spoke at StartupBREW Fargo on Wednesday, Sept. 9, 2026, at the Skyway Lounge in the Black Building. The Emerging Prairie event brings founders and operators together over coffee. Both Mangnall and Oelrich are University of Jamestown alumni building Credasis from North Dakota.

VERITY is Credasis’s instrument for label-free measurement: it calibrates on a system’s own normal behavior, then measures whether live data still matches that baseline. This allows systems to detect attacks, hallucinations, and bad hardware in real time, instead of after the loss of money, time, lawsuits, or data breaches. It is designed to run small, on ordinary CPUs, and stay local.

“We specialize in helping our clients make important decisions in an environment of uncertainty,” said Marty Oelrich. “People in high-stakes situations like medical, legal, and finance need to know that the systems they are working with aren’t under attack, hallucinating, or failing, so they can make those big decisions with systems they can trust.”

Credasis currently has prototypes in ICS (industrial control systems), brain-computer interface, network security, and quantum verification, and is developing a new AI encoder built around security and fidelity, designed to avoid inventing answers when evidence is thin.

Credasis is seeking design partners and research collaborators who work with unlabeled, high-stakes data.