Welcome to Google Parfait! The goal of this organization is to host repositories for defining and executing machine learning and analytics computations and workflows that enable strong privacy claims consistent with users’ privacy expectations, including:
- Libraries and frameworks used in research simulation
- Libraries and system components used in Google's production federated compute infrastructure
- Reference architectures for production systems
Our repositories support privacy-preserving technologies, including:
- Federated Learning
- Federated Analytics
- Differential Privacy
- TEE-enabled privacy architectures
- Personalization & Local Computation
- Privacy preserving aggregation, including approaches based on differential private and secure multiparty computation
- Private Information Retrieval (PIR)
Our technology and research has been used widely across Google, deployed by other tech companies, and referenced by many others. Some examples include analytics for Live Translate in Android’s Protected Compute Core, and Gboard’s Next Word Prediction and Out-of-Vocabulary discovery.
For those interested in production cross-silo or cross-device federated learning architectures that run on Google Cloud, please refer to the architecture in the Google Cloud Architecture center and their associated GitHub repository.
We are a team of researchers and engineers at Google that develop foundational technologies that enable strong privacy guarantees for AI and analytics systems, and maintain these repositories, which are used in production across various Google platforms. We believe that strong privacy guarantees are essential to the future of tech, and are committed to driving innovation in this space. We encourage engineers and researchers to actively engage with these repositories and use the code with confidence to build private-by-default systems and publish papers.
Want to learn more about us or get in contact? Visit http://federated.withgoogle.com.