Running Numpy programs homomorphically

Making FHE mainstream requires tools that make it easy to use for non-cryptographers. Being able to take regular Numpy programs and run them on encrypted data would enable data scientists to integrate FHE into their models, however complex they might be.

In this talk, the team at Zama will introduce a homomorphic numpy compiler, demonstrate how it can be used to run a recurrent neural network on encrypted data, and share their thoughts on homomorphic compiling.

# About the speakers:

Dr Rand Hindi is an entrepreneur and investor in deeptech. He is the CEO at Zama, and was formerly the CEO at Snips (acquired by Sonos in 2019). Rand is an investor in 30+ companies across cybersecurity, AI, blockchain, psychedelics and medtech. He holds a PhD in bioinformatics from UCL.

Samuel Tap is a doctoral researcher at Zama and INRIA, working on homomorphic compilation. He was the first hire at Zama, where he was originally working on cryptography research and the Concrete framework. If he wasn't a cryptographer, he would have been a luthier.

Ayoub Benaissa is a cryptography engineer at Zama focusing on homomorphic compilation. He was previously working with OpenMined, where he developed TenSEAL, a python library for homomorphic encryption.