Wodan AI is a privacy-preserving computation platform that converts standard artificial intelligence models into encrypted runtimes using homomorphic encryption (HE). The platform combines cryptography with containerized orchestration to facilitate data-in-use encryption, ensuring that information remains protected throughout ingestion, processing, and output. It supports the complete data lifecycle by managing client-side encryption and key storage locally while enabling secure remote inference.
The system operates by intercepting cleartext requests via a client deployed within the data owner's infrastructure, encrypting the data locally, and transmitting it alongside evaluation keys to a server for computation. The server processes the encrypted workloads and returns encrypted responses to the client for local decryption, facilitating secure applications such as machine learning, computer vision, rule engines, and encrypted databases without exposing sensitive or raw information.
