Ground every output
Important answers should remain connected to the evidence behind them.
Technology is for people.
San Francisco, California
What we believe
The people who know how things are made should spend their days deciding, not reconciling.
In every company there are people whose judgment is the actual product, and most of their week goes to reconciling documents. We build to give that time back: to inventing a process, judging a trade-off, teaching the next engineer, and talking to the customer. We start where the paperwork is heaviest, quoting engineered products and systems, and we go wherever expertise meets paperwork.
Why now
AI can finally work across the package. Trust still has to be designed in.
Modern models can interpret more of the technical and commercial material behind engineered work. But useful enterprise systems need more than generation. They need source evidence, domain structure, explicit uncertainty, access controls, and people who remain accountable for the result.
How we build.
Keep experts in control
AI should accelerate judgment, not hide or replace it.
Build for real complexity
Variants, revisions, exceptions, and incomplete information are the workflow, not edge cases.
Earn trust through the product
Security, traceability, and review should be visible in how the system works.
Built by operators who have worked both sides of the problem.
Maddie Zheng
Co-founder & CEOLed the Pixel A-series program at Google from concept to launch. Built AI infrastructure for large-scale models at Meta Superintelligence Labs, including trust and safety systems running inline at production scale. Dropped out of Harvard Business School to build Nyotta AI.
Yiyuan Zhang
Co-founder & CTOLed NLP and AI projects at Google Assistant, now part of Google DeepMind. Built enterprise search at Moveworks for Northrop Grumman, CVS, and other enterprises with the most fragmented legacy systems and highest security requirements. Graduate degree in CS and NLP from Yale.
