The platform behind governed, cross-institutional model training

Chefoba runs the federation, the governance, and the audit trail, so your team doesn't have to build any of it in-house.

How federation works

Your data stays with you. Only model updates move.

Chefoba runs a federated learning process across your institution and your partners. Each organisation trains a local model on its own data, inside its own environment. Chefoba's coordinator combines these local model updates, not raw records, into a single shared model that reflects patterns across every participating institution.

No participant ever sees another participant's transactions, customer records, or underlying data. What moves between institutions is model parameters: numeric summaries of what a local model learned, not the data it learned from.

The result behaves like a model trained on the combined dataset of every institution, without any institution actually pooling its data. In our own evaluation, this cost effectively nothing in predictive performance against a centralised approach on the same problem. The full methodology is on the Technology & Research page.

Institution A local model Institution B local model Institution C local model Institution D local model Chefoba trust-weighted aggregation shared model + audit trail

Model updates flow in. Raw data never does.

Core modules

What runs on top of the federation core

Module 01

Trust scoring

Every participating organisation is scored on the reliability and quality of its contributions over time. Chefoba tracks how consistent, well-formed, and useful each institution's model updates are, and surfaces that score to administrators. This is a proprietary capability; none of the platforms we track offer it.

Module 02

Adaptive weighting of model contributions

Trust scores feed directly into aggregation. Institutions with a stronger track record, or with data more representative of the problem at hand, have their updates weighted accordingly, rather than every participant counting equally regardless of data quality or consistency.

Module 03

No-code policy management

Compliance and risk teams set participation rules, aggregation policy, and data-handling constraints through a policy interface, not through engineering tickets. Policy changes are versioned and auditable.

Module 04

Automated regulatory evidence generation

Every training round, aggregation decision, and policy change generates evidence automatically: what happened, when, under what policy, and with what result. This evidence is structured for compliance review and regulator inquiry, not assembled after the fact.

Deployment

Cloud, or private cloud and on-premises

For regulated banking buyers, where the platform runs is often as important as what it does.

Cloud

Fully managed deployment for institutions that want to move quickly and don't require data residency guarantees beyond standard regional hosting controls.

Private cloud & on-premises

For banks and fintechs under stricter data residency, outsourcing, or regulatory requirements, Chefoba deploys inside your own private cloud or on-premises infrastructure. This is called out explicitly because it's a real differentiator: several federation-native tools in this space are cloud-only or framework-only, with no equivalent managed on-prem path.

Be first to know when Chefoba opens access.

We're onboarding a limited group of fintech and banking partners before general availability.

Join the waitlist