Skip to content

Introduction🔗

Apheris is a federated Computational Governance Solution for data-driven analytics. It provides all the building blocks to enable and automate secure, governed and privacy-preserving computations on data.

Main capabilities of the Apheris product are:

  • Data residency: computations are brought to data so it doesn't have to move
  • Computational Governance: full control and governance of computations on algorithmic level
  • Scalability: single or multiple workloads on an arbitrary number of remote datasets. Any data and any data driven-algorithm can be used.
  • ML compatibility: ML framework agnostic (PyTorch, Tensorflow, Scikit-Learn, etc). Via the integration of NVIDIA Flare, Apheris provides an interface to port any ML model from e.g. MONAI, BioNeMo, Hugging Face or other sources with minimal effort.
  • Enterprise-ready federated learning: The Apheris central orchestrator runs on a managed, security-hardened infrastructure and provides industry-standard features for enterprise-grade usage. (if your use case requires a different infrastructure setup, contact us via support@apheris.com)

Simply said, Apheris provides all necessary capabilities to build scalable collaborative data ecosystems with and for anyone.

Governance, privacy and security for federated learning🔗

The key product of Apheris is the Compute Gateway which runs within a data custodian's environment. It is a lightweight component to validate and securely execute computations based on defined policies. Compute Gateways have a small footprint - easy to deploy into various environments.

Apheris is based on a federated architecture, using the open source NVIDIA Flare engine to enable the federation and controlled execution of analytics and machine learning algorithms. Based on this architecture, any number of Compute Gateways and datasets can be used within federated computations.

Apheris is built to enable safe and compliant collaborations between data providers and data consumers.

Apheris Federated Governance Solution Overview

For Data Custodians🔗

After deployment of the Compute Gateway, data custodians can access the Governance Portal - an intuitive UI to control all tasks needed for ensuring security, privacy and compliance during collaborations:

Apheris Governance Portal Homescreen

Information Security professionals will find detailed information about the security posture, architecture and risk mitigation strategies during collaborations within the Trust Center together with latest penetration test results and more.

For Data Scientists🔗

Data Scientists can stay in their preferred workflows, tools and MLOps setup. Apheris provides a complete CLI to interact with the Apheris product.

Via the CLI Data Scientists and ML Engineers can

  • access metadata of a dataset and details about requirements for acceptable computations
  • specify computations via Compute Specs
  • select models running out-of-the-box from the Model Registry
  • simulate and run statistics and machine learning workflows
  • and assess the status of submitted compute jobs

Workflow with Apheris🔗

For all partners within a collaboration, Apheris provides a straight forward workflow.

Workflow Data Custodians and Data Scientists within Apheris

Data custodians set the requirements for computations on a registered datasets. Data Scientists create compute specifications which need to comply with defined requirements.

Apheris ensures the

  • automated validation of submitted computations
  • encrypted transmission of computations and results
  • aggregation of results
  • detailed logging of interactions with registered data
  • and data residency

Sensitive data will never leave the data custodians premises but can be safely contributed to collaborations.