Chefoba started as academic research into whether federated learning could genuinely work for cross-institutional fraud and AML detection, not as a product idea looking for a research story to attach to it.
Origin
The federation core behind Chefoba began as academic research evaluating whether federated learning, differential privacy, and secure aggregation could work together on a real cross-institutional financial risk problem, without the usual gap between a research paper's claims and how the approach behaves in practice.
That research was carried out on real remittance transaction data across two UK fintech platforms, and it was designed from the outset to report results honestly, including the trade-offs and limitations of privacy-preserving techniques, rather than presenting only the results that supported a clean narrative. The corridor-level fairness gap found during that evaluation, and the finding that apparent fairness gains under differential privacy can be an artefact of degenerate calibration, are both things a less careful evaluation could easily have missed or glossed over.
Chefoba is the product built on top of that research: the governance, trust scoring, adaptive weighting, and regulatory evidence generation that a research prototype doesn't need, but a production deployment inside a regulated institution does.
Get in touch
hello@chefoba.xyz
For the fastest response, use the waitlist form and tell us your use case.
We're onboarding a limited group of fintech and banking partners before general availability.
Join the waitlist