~/startup-journey.md

Startup journey

In 2020, Ilya Razenshteyn and I founded CipherMode Labs in Los Angeles to turn my PhD research into a product. The goal was simple to state and hard to build: let companies compute on their data while it stays encrypted, so they can use and share even their most sensitive information without ever exposing it. We later renamed the company Pyte, short for "Private Byte". In February 2025, Pyte was acquired by All Hands AI.

The problem

Companies sit on valuable data they can't fully use. Privacy rules, security risk and plain distrust keep it locked away, and the usual ways of sharing it (copying it into one central place, or masking it) either create new risks or destroy what makes the data useful.

Our answer was secure multiparty computation (SMPC): several parties compute together on their combined data while every input stays encrypted the whole time, including during the computation itself. The math had existed for years, but it was slow and only usable by cryptographers. We set out to make it fast and usable by any engineer.

What we built

  • CipherCore (2022), an open-source library that made SMPC usable without any cryptography knowledge. Developers wrote programs in their normal workflow and CipherCore turned them into secure protocols.
  • SecureMatch (2023), on-premise data collaboration: two or more companies could measure how much their customer lists overlap without the data ever leaving their own cloud. A Fortune 100 consumer goods company used it to evaluate potential data partners.
  • SecureSuite (2024), the broader platform for matching, joining datasets and building AI models on encrypted data. The platform was SOC 2 compliant.

The team had six PhDs, three of them experts in secure computation, and three ICPC gold medalists, with offices across North America and Europe.

Timeline

Talks from those years, including at the AI4 conference and MIT, are on the talks page.

Press