Federated AI with Zero data movement.

Cross-boundary analytics without crossing any lines.

Brane is a federated orchestration framework built for privacy, trusted by research institutions and enterprises. Package your algorithms as containers, define workflows in BraneScript, and run distributed applications across organizational boundaries without moving sensitive data.

Trusted byUniversity of AmsterdamApache 2.0 LicensedGDPR CompliantEU Funded Research

The Collaboration Paradox.

You need joint insights, but data sovereignty and regulations (GDPR, HIPAA) keep your most valuable data siloed.

Data Sovereignty

Sensitive datasets must remain within their organizational infrastructure. Brane ensures that raw data never crosses organizational boundaries.

Regulatory Compliance

GDPR, HIPAA, and national security laws restrict data movement. Brane supports compliance by design -- computation moves to the data, not the other way around.

The Trust Gap

Collaborate without exposing raw data to partners. Brane enables privacy-preserving joint analysis where only approved results are shared.

Computation moves. Data doesn't.

Package

Containerize your algorithms and data processing steps as portable, reusable building blocks that run anywhere.

Orchestrate

Write simple workflows in BraneScript to define what computations happen, and where they run across the network.

Protect

Data governance policies automatically ensure that raw data never leaves its source. Only approved computations execute locally.

Privacy is not a feature. It is the architecture.

Structural Privacy

Privacy isn't a setting -- it's the architecture. Brane moves computations to the data, ensuring raw datasets never leave their source.

Programmable Policy

Define fine-grained, code-based access policies that govern exactly what computations can run on your datasets.

Intuitive BraneScript

A developer-friendly DSL for writing distributed workflows. No complex distributed systems expertise required.

True Federation

Connect multiple organizations into a trustless network while maintaining independent control over local data.

Vendor Agnostic

Algorithms are packaged as standard containers. Run on any infrastructure that supports the Brane runtime.

Open Source Core

Auditable, extensible, and transparent. Deploy with confidence under a permissive open-source license.

Regulated industries. Distributed data. One framework.

Federated Clinical Research

Hospitals and research institutions can collaboratively train machine learning models across patient data without any records leaving the hospital. Brane supports multi-site clinical trials, rare disease research, and population health studies while keeping data within institutional boundaries.

  • Multi-hospital model training for disease prediction without sharing patient records
  • Collaborative drug discovery across pharmaceutical companies and academic labs
  • Cross-border epidemiological research while complying with national health data regulations

Describe the workflow. Brane handles the distribution.

workflow.bs
import train_model;
import aggregate;

// Train locally at each site
let model_a := train_model(new Data{ name := "site_a_data" });
let model_b := train_model(new Data{ name := "site_b_data" });

// Aggregate without sharing raw data
let result := aggregate(model_a, model_b);

println(result);
What happens under the hood

Import packaged algorithms

Pre-built container packages for training and aggregation are imported into the workflow.

Train at each location

The train_model function executes at the site where the data resides. No raw records cross any boundary.

Only models travel

Trained model parameters (not raw data) are passed to the aggregation step. Policies verify this at every boundary.

Combine results securely

The aggregate function combines model outputs into a final result. Governance rules are enforced end to end.

Three sites. Zero data movement.

Brane Orchestrator

Site A

data stays here

Site B

data stays here

Site C

data stays here

Brane sends containerized computations to each site, where they execute against local data. Only approved results travel back -- raw datasets never cross organizational boundaries. Policy enforcement happens at every step.

Deploy your first federated workflow today.

Everything you need to get started is in the docs. Set up a federated environment and run privacy-preserving computations in minutes.