Better Together – Duality Adds Native Support for Flower for End-to-End Federated Learning

Federated learning has grown into a vibrant ecosystem with frameworks that excel at different stages of the journey from experimentation and prototyping to large-scale deployment across real organizations. As this ecosystem matures, interoperability has become mission-critical. Teams want the freedom to innovate with the tools they love while ensuring they can deploy confidently in governed, production environments.

Today, we’re excited to share that Duality is adding native support for Flower, one of the most widely used and developer-friendly federated learning (FL) frameworks. This integration strengthens the broader FL ecosystem by enabling Flower workloads to run inside Duality’s secure, policy-driven orchestration layer without code changes or workflow rewrites.

A Complementary Ecosystem: Flower, FLARE, and Duality 

Federated learning now spans a range of frameworks, each designed with a different purpose in mind:

• Flower offers a highly flexible, Pythonic interface ideal for rapid experimentation, iterative development, and custom FL logic across diverse environments.
• NVIDIA FLARE provides powerful orchestration and deployment capabilities for organizations running scalable, production-ready FL workloads.
• Duality acts as the unifying platform providing governance and privacy layer, ensuring that regardless of where the FL workflow originates: Flower or  FLARE, or other open-source frameworks it can run securely, compliantly, and at mission scale. It also combines additional privacy technologies (e.g. TEEs) on top of the federated computation. 

These layers don’t compete; they complement one another. Flower accelerates innovation. FLARE scales deployment. Duality ensures governance, control, and PET-based security across data owners and jurisdictions. Furthermore, when the model training is done, Duality also provides the ability to run secure inference using the trained model.

Together, they bring the full FL lifecycle, from prototype to production, into a coherent, secure, and interoperable environment.

Running Flower Workloads Securely at Scale

Teams that prototype with Flower can now deploy those same workloads directly inside the Duality Platform. No refactoring. No new orchestration logic. Just bring your existing Flower code and Duality handles:

• Governance & Policy Enforcement
  Define participation rules, data-use constraints, and model-to-data mapping across organizations.
• Privacy-Enhancing Technologies
  Add cryptographic protection to intermediate weights and run aggregation inside secure execution environments for stronger privacy guarantees.
• Scalable Multi-Party Orchestration
  Manage distributed training jobs across hospitals, banks, agencies, and partners with built-in lifecycle automation.
• Compliance & Auditability
  Enforce regulatory and organizational standards without modifying core FL code.

This approach lets teams use Flower’s familiar API, with Duality ensuring that the resulting workflows meet enterprise and public-sector requirements.

Why This Matters

Organizations no longer need to choose between:

• the flexibility of open-source FL frameworks, and
• the rigor of enterprise-grade governance and security.

Now they get both.

A model can be developed in Flower, orchestrated through FLARE, and governed and secured end-to-end by Duality, all as part of a single, interoperable stack.

The result:

• Faster transitions from research to deployment
• Reduced duplication across tools
• Consistent policy enforcement
• Strong PET-based safeguards
• Zero data movement outside organizational boundaries

This is the path toward scaling federated learning in real-world environments across industries, countries, and organizational lines.

Looking Ahead

Native support for Flower is another step toward Duality’s mission: building a secure, interoperable platform for collaborative AI, one that embraces the strengths of the entire FL ecosystem rather than replacing it.

As we deepen integrations with frameworks like NVIDIA FLARE and expand support for PET-enabled workflows, Duality is creating a unified environment where innovation thrives and compliance holds firm.

If your team is building federated learning pipelines, whether in Flower, FLARE, or custom FL code, and you’re ready to scale securely, we’d love to connect.

Oded Rosolio Senior Data Scientist, Duality Technologies