The AI Framework team is building an open source framework for homomorphic machine learning. The product your team will design, develop, test and maintain aims to take any DAG (directed acyclic graph) —- or later, any program —- and to transform it into a fully homomorphic equivalent, with tunable and small accuracy loss. Notably, a first usage will be to convert neural networks into end-to-end encrypted equivalent programs, which can virtually run on an insecure server while maintaining user’s privacy. In this manner, the product has strong relations with compilers or code-to-code translators. Challenges are about having good performance, small accuracy loss and no compromise on security at the same time.

Your team (and thus you) will be responsible for:

• turning any program in a fully homomorphic equivalent (with the help of our crypto research team)
• finding the best parameters following the constraints, to optimize speed, RAM & data size, while maintaining security
• developping the product in a clear & efficient way, mainly in python
• making the use of our compiler as easy to use as possible for our users
• publishing or blog the techniques used by the tool
• managing our open source repositories and the community around it

• experience implementing compilers or code-to-code translator
• development of production tools in python or C/C++/Rust
• optionally: experience implementing optimization passes or operational analysis
• optionally: experience with FHE or cryptography
• optionally: experience with deep learning
• be passionate about privacy and open source

Our process is described in detail here: https://zama.ai/2020/04/28/how-we-hire-at-zama/