Why we founded Cull.
To fix what record-everything AI got wrong. To protect the privacy of the people being recorded. And to earn the trust that lets AI and robotics truly help humanity.
Cull.ai exists because today's machines listen first and ask permission never, and real people pay for that: the guest on the call who never agreed, the patient whose disclosure becomes a record, the family whose kitchen conversation ends up on a server.
“The people being recorded deserve a say in what a machine keeps about them.”
Privacy is not friction to be engineered away. It is the foundation that trust is built on.
And trust is the whole prize. The future we are building toward, assistants you can speak to candidly, robots that care for aging parents, machines that help in hospitals, homes, and schools, arrives only if people can welcome a machine into the room without fear. Our mission is to make machines worthy of that welcome, and everything on this website, the architecture, the patents, the hiring, serves it.
Borrowed deliberately from the great AI safety labs: a small team whose prior work is the credential, a thesis argued in public, capability and governance engineered together from the first commit, and staged capital raised against verifiable milestones. At this stage the team is the product, and this website is a recruiting page on purpose.
Our founding CTO is Hitesh Anand Gupta, employee #10 at Otter.ai for eight years. He built the system that tells speakers apart, the system that recognizes each voice, and the harder streaming versions that do both live. Twelve of the nineteen Otter patent filings in the public record name him as an inventor. The one problem our product lives or dies on, knowing who is speaking before anything is written, is the problem he already solved at production scale.
We are incubated by PatentVC, the venture studio of LegalForce RAPC Worldwide, the IP firm behind Trademarkia, funded from the operating profits of a platform with more than 125,000 customers. A voice-AI lab born inside a law practice: privacy law and patent strategy in the founding DNA rather than retrofitted before a Series B.
Six accelerated Track One filings span the governance stack, drafted apparatus-generic so the same claims travel from a bot in a video call to a robot in a kitchen. The parent was on file before the consolidated litigation reshaped the category. Priority dates are the one asset that cannot be bought later, and the claims were drafted by counsel who has spent two decades enforcing patents in federal court. Everything is built clean-room: fresh hardware, logged provenance, original work only.