AeonX HML Platform
Federated Homomorphic ML Infrastructure for Secured ML Model Training
AeonX HML is an infrastructure that allows training machine learning models using data from multiple isolated sources without exposing raw data or models. It solves critical security blockages in regulated industries such as Healthcare, Finance, and Space & Defense.
How Federated Homomorphic Training Works
AeonX HML uses Model Homomorphic Computation methods. The platform allows you to train ML models securely and collaboratively without exposing raw training datasets to external parties or hosting servers.
All AeonX Team ML servers run directly within your local secure infrastructure. The central AeonX Homomorphic Computer server does only the homomorphic computation and synchronization of parameters. Zero raw training data ever leaves the local team servers.
Decentralized Federated Training
Manage training processes and teams using an intuitive web interface. Each computed node is set up on-premise. Whether using a laptop or a high-performance computer cluster, you can train and monitor model health without sharing raw inputs.
Cloud & On-Premise Agility
AeonX HML can be utilized through the AeonX Homomorphic cloud services or installed completely on-premise. Organizations can license the software suite to deploy it inside their preferred, air-gapped infrastructure.
Training Workflow
Integrate seamlessly with standard ML development workflows and popular open-source libraries.
Design Model
Design models locally using TensorFlow or PyTorch libraries.
Upload Architecture
Upload the model architecture definitions to the secure HML server.
Deploy Node Servers
Run local Team ML servers in your isolated data boundaries.
Synchronize
HML server automatically computes updates homomorphically.
Target Industries
HML makes machine learning possible in heavily regulated environments.