Open source federated learning software

The most proven federated learning software designed for healthcare research

Aggregate models and insights - not data

Substra is a ready-to-use, open source federated learning (FL) software developed by Owkin, now hosted by the Linux Foundation for AI and Data.

Substra enables the training and validation of machine learning models on distributed datasets. It includes a flexible Python interface and a web application to run federated learning training at scale.

Academic research centers and biopharma companies deploy Substra (formerly Owkin Connect) in a wide variety of federated learning settings for clinical research, drug discovery and development.

Federated learning

Why use Substra?

Data agnostic

Can be used with many different data modalities such as video, images, audio, time series or tabular data.

Framework agnostic

Compatible with any machine learning model from any framework including TensorFlow, Torch and Scikit-Learn.

Infrastructure agnostic

Works seamlessly with any IT infrastructure setup - on premise or in the cloud.

Flexible

Can be deployed in a federated production environment or used on a single machine to perform simulations and debug code.

Traceable

Monitoring, traceability and fine-grained permissions for complete control of data manipulations.

Secure

Proven in strict privacy settings to meet global compliance requirements for hospitals and pharmaceutical companies.

01/04

It all starts with data

A data scientist registers their data on their Substra servers for machine learning model training.

02/04

Compatible with any model

A data scientist registers a model to be trained on Substra.

03/04

Secure model training

A data scientist trains the model, with computation done securely and autonomously through the Substra network.

04/04

Easily monitor and trace results

Everyone can examine and visualise model performance at the end of the training process to select the best option.

FAQs

Substra in action

Federated drug discovery

Largest ever pharma-industry AI collaboration for federated drug discovery.

Substra securely connected 10 pharma partners supported by 7 tech partners, for collaboration between 100+ experts.

See results

Federated drug discovery

Largest ever pharma-industry AI collaboration for federated drug discovery.

Substra securely connected 10 pharma partners supported by 7 tech partners, for collaboration between 100+ experts.

See results

Biopharma

The MELLODDY project

Collaborative research project

The HealthChain project is a collaborative research project between Owkin Federated Data Network partners using Federated Learning.

Machine Learning models were trained on real histology slides and dermoscopy images across for institutions to predict treatment response in breast cancer and melanoma patients. The resultant model helps oncologists better predict treatment response in breast cancer.

Read paper

Collaborative research project

The HealthChain project is a collaborative research project between Owkin Federated Data Network partners using Federated Learning.

Read paper

Clinical research

The HealthChain project

Substra

Discover how to use and deploy Substra in a real-world setting.

Voice as biomarker

Owkin and 12 partners are using Substra to train machine learning models on the largest distributed clinical dataset of voice data to identify biomarkers from voice data and create low-cost diagnostic tools that will be used alongside other clinical methods.

Substra

Product documentation