Skip to main content

Introduction to Bitfount

Welcome to Bitfount! Our mission is to make the world’s intractable data interactable. Our platform can help you tackle your most challenging problems at the intersection of data science, data collaboration, and data privacy.

Why Bitfount?

Collected data is often vastly underutilised due to barriers to collaboration stemming from privacy, security, or commercial sensitivity of the datasets, as well as technical barriers to the creation and maintenance of multiple up-to-date copies of a dataset.

Bitfount remedies this by providing a flexible, easy-to-use platform for privacy-preserving data collaboration. It enables data teams, analysts, and researchers to gather statistical insights as well as train and evaluate machine learning models on data which they don't have access to in raw form, whether internal to their organisation or external. It also enables data custodians to share the benefits of their data without giving up control or privacy.

Bitfount is built on the concept of Federated Data Science, which enables a data scientist or analyst to send encrypted code to data where it already lives and retrieve the results of analysis queries or machine learning tasks without the need to access raw data. This means data custodians no longer need to transfer data or take on unnecessary risk to collaborate with internal or external data science talent.

Bitfount can be accessed via a no-code desktop or web application, and via an extensible open source SDK.

How it works

You can think of Bitfount as the infrastructure connecting all parties involved in your data collaboration. Bitfount enables analysis of data at its source, and also contains a powerful usage-based access control layer which can be configured to govern the way in which users collaborate.

There are three key user types on the platform:

  • Project owner: user responsible for convening collaborators, setting the project's terms & conditions, and selecting the data science task to be carried out within the project.
  • Data scientist: user who creates AI models and data science tasks for use within one or more projects. Tasks can include machine learning algorithms, analytics queries and other analysis types.
  • Data custodian: user connecting a dataset and linking it to the project for analysis. No data is transferred away from the data custodian's systems.


Uses of the Bitfount platform

The Bitfount platform can be used in multiple ways:

  • Collaborative inference: Deploy existing AI models on federated datasets without requiring any data upload. For example, a clinical trial sponsor may wish to work with clinic sites to run medical AI models against patient data to improve clinical trial patient recruitment outcomes.
  • Federated learning: Train and fine tune AI models on federated datasets, from LLMs and other foundation models, to image classification and segmentation, and everything in between. For example, train a model designed to classify diseases from medical scans taken and held securely at multiple different health institutions.
  • Internal use: Secure data collaboration across jurisdictions or departmental silos. For example, retailers may wish to enforce privacy-preserving techniques against loyalty card data when their internal data scientists query the data.
  • 3rd party data evaluation: Evaluating third party data while maintaining privacy and control. For example, a data buyer may wish to obtain summary statistics on a third party dataset prior to purchasing it for research or model-training purposes.
  • Private Set Intersection (PSI): Understand the overlapping records between your dataset and another party’s without acquiring any knowledge of the records which aren’t in the overlap (intersection). For example, in dataset procurement, evaluate the extent to which your data overlaps with another before making a purchase of the raw data.
  • Evaluation of AI models: Evaluating models on distributed datasets without centralising data. For example, a regulator may wish to audit the effectiveness of an online trust and safety model without physical transfer of harmful content material. Or, an enterprise may easily evaluate a selection of open source models to understand which performs best on their internal data.
  • Modeller consulting: Outsource ML model development without granting access to raw data. For example, a financial services institution may wish to improve its automated fraud detection capabilities without needing to hire an in-house ML team.
  • Embedded services: Improve your products by training on your customers’ on-premise data. For example, a clinical research organisation may wish to embed Bitfount in their clinical trial recruitment software as part of their patient recruitment suite of products.

Next steps

Now that you have a better understanding of Bitfount and how it works, you're ready to learn about Accessing Bitfount.

Need Help?

If you have any questions after reviewing our Guides and Tutorials, visit our FAQs. If you can't find the answer you are looking for or would like to discuss anything further, please contact us at We're here to help!