The Helix product
Four components. One privacy-preserving signal.
An Android SDK, an iOS SDK, an on-device anomaly model trained to each user's signature, and a UK-hosted control plane your risk team operates.
Android SDK
Kotlin. Distributed via Maven Central. Targets API 24+. Less than 10MB on disk including the on-device model. One engineer-day integration.
iOS SDK
Swift. Distributed via Swift Package Manager. Targets iOS 15+. Same integration profile, same size envelope, same model.
On-device anomaly model
TensorFlow Lite (Android) and Core ML (iOS). Trained per-user across 10–15 sessions. Parameters held encrypted in Android Keystore / iOS Secure Enclave.
Control plane
Multi-tenant SaaS on AWS eu-west-2 (London). Configure policy, review aggregated risk events, integrate via webhook or REST. UK GDPR-compliant by design.
On-device by design
Most mobile-fraud SDKs forward raw signal to a vendor cloud. Helix runs the analysis on the device and forwards only the score and structured metadata.
Continuous, not transactional
Helix scores every interaction with the host app — not just the moment of payment. You see behavioural drift before fraud, not after.
Mid-market priced
From £15,000 annual platform fee plus per-MAU consumption. Accessible to a 100K-MAU challenger; scales to a 5M-MAU incumbent without rearchitecture.
See it in your app.
Become a design partner — production access, white-glove integration support.
Become a design partner