AI Potluck
Contribute
Back to Gap Map Model components / Federated learning

Apheris Networks

Apheris
closed / Overall score: n/a

Apheris's federated networks and the Compute Gateway that runs them. Each data custodian deploys the Gateway into its own environment from Apheris's Helm chart; the Gateway polls the Apheris orchestrator, checks every computation against the custodian's asset policies and privacy controls, and runs federated statistics or federated training, with NVIDIA FLARE as the main federation framework. Apheris runs industry networks for drug discovery on it, such as the AI Structural Biology network of pharmaceutical companies.

Closed, and on the map as an ADR-005 best-in-class comparator for this category. The installable `apheris-cli` package is a client for submitting jobs to the product, not the product, so it is not declared.

Openness

1 high confidence
1.0
license
Proprietary(closed platform
source
closed(no published implementation

Apheris publishes no implementation: data custodians install the Compute Gateway from a private registry using credentials Apheris issues, and the orchestrator it connects to is Apheris's own service.

Adoption

not assessed

Apheris publishes network and dataset counts, such as nine pharmaceutical companies in one network, but no count of deployments or users, and it has no public distribution channel that could be counted, so no adoption level is recorded.

  • https://www.apheris.com/networks/aisb recorded 2026-09-27

    The AI Structural Biology network: "An industry-led collaboration of nine leading pharmaceutical companies". A membership figure for one network, not a usage count for the platform.

Capability

4 medium confidence
4.0

Every data custodian operates its own gateway, and every computation is checked against that custodian's policy before it runs, which matches Flower's production reach. Training runs on NVIDIA FLARE, and Apheris documents no reach beyond the data center.

Verified 2026-09-27