platoseedplatoseed
End-to-end encrypted AI.
Blyss is an AI model provider built on strong security guarantees. We run AI workloads in GPU-based secure enclaves, which keep data encrypted even while it's being processed. Our customers get a privacy model similar to on-premises deployment, without ever having to think about AI infrastructure or GPU supply. The Blyss team has strong experience in cryptography and AI from Stanford, Apple, Nvidia, and more. We think data privacy is fundamental to safe, trustworthy AI, so it's our job to make end-to-end encrypted AI a no-compromise option - just like Signal and WhatsApp did for messaging.
Worked at Yubico. Wrote ia.cr/2022/368. Stanford CS '19 BS + MS.
I've spent my whole career making computers go fast, especially for deep learning. I love large scale compute and care deeply about democratizing this power. Previously, I worked on the Apple Neural Engine, and NVIDIA's cuDNN. Stanford '19, BS and MS in EE.
We make it easy to use and deploy homomorphic encryption.
Blyss offers a developer-focused API for private retrieval in a cloud storage-like key-value store, enabling cryptographically private data access so that neither the service nor providers learn which items are retrieved. The launch positions Blyss as a privacy tool for developers, highlighting its SDK, use cases (password checks, private DNS, code scanning), and a cloud storage-like service that handles encryption and privacy automatically.
From the original launch (Feb 2023) — may be outdated.
Formerly “Spiral” · why startups rename →
The enterprise platform for AI phone calls
Encrypted AI with verifiable privacy
