Any data, any size, anywhere
Want to learn more about how Apheris can help you power your infrastructure with federated machine learning and analytics?
Apheris plugs into your existing infrastructure and integrates with other elements of your ML stack. Use existing infrastructure for data and compute, empowering you to use a variety of model training and serving tools – all across organizational boundaries.
The Apheris Compute Orchestrator sits on top of multiple federated data sources and orchestrates ML and analytics across the different organizations. Only aggregated results, and not raw data, is returned to the user.
Data custodians are in full control over what happens with their data. They define who can run ML and analytics on their data, and which privacy controls should be applied, while ensuring that results are returned to the user without exposing data or IP.
Data is often spread across different departments or organizations and resides in a variety of systems. Access these for federated ML and analytics without the need to move data. This protects IP, ensures data privacy, and eliminates the costs of centralizing data.
Each data custodian deploys the Apheris Compute Gateway within their existing infrastructure (either cloud or on-prem) and connects it to the data. Apheris provides abstractions for the data sources so users don’t need to focus on this. The data custodian controls which specific computations they allow to run on their data, without their data ever leaving their premises.
The Apheris Compute Gateway empowers external parties to run computations on the connected data. It includes a compute environment from Apheris, based on Python/Kubernetes, that contains statistics and machine learning modules designed for federated computations. Alternatively, you can connect the Compute Gateway to external compute environments, such as Spark, Slurm, etc.
With the Apheris SDK you can use a variety of machine learning tools, frameworks, and libraries. Bring your existing data pipelines and models, inclunding pre-trained models, and run them with simple glue code in the Apheris platform.
The SDK preserves your existing code and repo structure, so that you can easily version control the glue code file(s) together with your existing code using Git.
The Apheris SDK includes dedicated ML and data science modules for privacy-preserving statistics and ML, for example, and empowers you to do foundation model fine-tuning and more.
Data custodians are in full control over what happens with their data – they define asset policies with privacy controls that specify which user is permitted to run which computation on their data.
For any computation request from a user, these asset policies are validated throughout the platform. Additionally, data custodians can approve incoming compute requests and control what results leave their Compute Gateway. This ensures that results are returned to the data scientist without exposing data or IP.
Apheris cleanly interfaces with downstream tools that leverage the analysis results or trained models. You can simply export results to use them in your data application, and also pass on trained models for model serving in other systems that are already used in your organization.
"With Apheris, we found the perfect partner for exploring privacy preserving data analytics and optimization along the value chain with our customers."
Want to learn more about how Apheris can help you power your infrastructure with federated machine learning and analytics?
DevOps and MLOps are common methodologies in every company that wants to become software and data science driven by weaving AI into the core fabric of their business. Read what is required to securely collaborate with partners on data and AI at scale.
Learn how to leverage federated and sensitive data at scale and how you can collaborate with your partners to advance and evolve the practice of life sciences.
If you’ve been successful to date with building industry-leading data products, you don’t want to rest on your laurels. Instead, you should be trying to extend your lead. However, this can be daunting task in an industry that is in constant flux as the data, machine learning and AI landscape.