Cifer

Founded 2023

Train and deploy machine learning models across institutions without sharing private raw data

You can use Cifer Workspace to train and deploy machine learning models collaboratively across institutions without sharing sensitive raw data. It combines federated learning with fully homomorphic encryption to perform computations directly on encrypted datasets. Through a no-code interface or Jupyter notebooks, you can configure training rounds, manage participating nodes, and orchestrate experiments. The system logs all model updates and contributions to an immutable audit trail using blockchain provenance, letting teams collaborate safely on confidential projects in finance, healthcare, and research.

What Cifer does

Configure federated training rounds and node connection settingsRun machine learning computations on fully homomorphic encrypted dataReview and approve model weight updates before aggregationDeploy federated nodes across private Kubernetes clustersLog training participation and model weights to a tamper-evident audit trailConnect Jupyter notebooks to remote decentralized training nodesDecentralized model training across institutions using federated learning without sharing raw datasetsEncrypted computation during training and evaluation using Fully Homomorphic EncryptionImmutable contribution logs and model version tracking with blockchain provenanceNo-code interface for configuring federated nodes, training rounds, and hyperparametersDirect integration with Jupyter notebooks for testing and experimental prototypingHuman-in-the-loop consent checkpoints requiring explicit approval for model updatesMulti-cloud and on-premises deployment options across Docker, VMs, or Kubernetes clusters