Apheris (apheris AI GmbH)
Executive Overview
Apheris (officially apheris AI GmbH) is a deep-tech company based in Berlin, Germany, that provides a world-leading platform for federated computing and secure data collaboration. Founded in 2019 by Robin Rohm and Michael Hoh, the company addresses one of the most significant challenges in modern research: the "data silo" problem. Apheris enables organizations to securely connect and analyze sensitive datasets across organizational, geographic, and regulatory boundaries without ever moving or sharing the raw data.
Apheris is a pioneer in the field of Federated Machine Learning, particularly within the pharmaceutical and life sciences sectors. By allowing AI models to travel to the data?rather than moving data to the models?Apheris protects intellectual property (IP) and ensures compliance with strict data privacy laws like GDPR. The company is a key player in high-stakes fields such as drug discovery, where it facilitates collaborative research between global pharmaceutical giants, biotech startups, and healthcare providers.
Continue…
Core Technology: Federated Computing and Governance
The Apheris platform is built on a "Compute-to-Data" architecture. This means that sensitive information stays behind the firewall of the data custodian, while only encrypted model updates or statistical aggregates are shared with the central orchestrator.
1. The Federated Orchestrator
The central component of the network is the Apheris Orchestrator, which manages the collaboration between multiple data providers. It coordinates the distribution of AI models, the aggregation of learning results, and the overall workflow without ever having access to the underlying raw data.
2. Computational Governance
Apheris provides an "End-to-End Computational Governance" solution. This framework ensures that every computation launched within the network is compliant with pre-defined rules.
* Asset Policies: Data custodians can define exactly what type of analysis is permitted on their data.
* Compute Specs: Data scientists submit specific "specifications" for their intended analysis, which must be approved by the custodian before execution.
* Audit Trails: Every interaction with the data is logged, providing a verifiable record for regulatory compliance and IP protection.
3. Integration with NVIDIA FLARE
In 2024, Apheris integrated NVIDIA FLARE (Federated Learning Application Runtime Environment) into its platform. This partnership combines NVIDIA?s scalable federation engine with Apheris's robust governance and security layers, allowing for production-grade AI training on a global scale.
Products and Services
Apheris offers a suite of tools designed to move federated AI from research concepts into industrial applications.
1. Apheris Compute Gateway
The Apheris Compute Gateway is a lightweight software agent deployed within a data provider's own IT environment (on-premise or cloud).
* Local Execution: It serves as the "execution engine" that runs the AI models locally on the data.
* Data Access Layer (DAL): The Gateway manages secure connections to various data sources, such as Amazon S3, Redshift, or local databases, while enforcing the custodian's governance rules.
2. Apheris Governance Portal
The Governance Portal is a web-based UI that allows data custodians to manage their participation in research networks. Through this portal, users can:
* Register and manage datasets.
* Configure Asset Policies and review Compute Specs.
* Monitor the progress of federated training rounds.
* Access detailed logs for auditing and security verification.
3. ApherisFold
Launched in late 2025, ApherisFold is a specialized product designed for the latest era of structural biology.
* Model Customization: It allows pharmaceutical companies to securely run, benchmark, and fine-tune advanced protein co-folding models like OpenFold3 and Boltz-2.
* Proprietary Data Leverage: Researchers can apply these models to their own proprietary structural data to improve accuracy on novel drug targets without risking the exposure of their most valuable discoveries.
4. Federated Data Networks
Apheris facilitates several flagship research networks:
* ADMET Network: Focused on predicting the Absorption, Distribution, Metabolism, Excretion, and Toxicity (ADMET) properties of small molecules to reduce clinical trial failures.
* AISB Network: The AI Structural Biology network, which focuses on protein co-folding and binding affinity prediction to accelerate the design of new therapeutics.
Strategic Significance and Partnerships
As of 2026, Apheris has become an essential infrastructure provider for the "Top 20" global pharmaceutical companies, including collaborations with Roche, Johnson & Johnson, and AbbVie. The company is ISO 27001 certified, reflecting its commitment to the highest standards of information security.
Through partnerships with AWS and NVIDIA, Apheris provides a "sovereign cloud" experience for life sciences, ensuring that even when using public cloud infrastructure, data remains strictly confined within the owner's virtual private cloud (VPC).
