MAYAS.AK
Back to workProduction SaaS

SyriaHuwiya

Enterprise eKYC identity verification platform with biometric face matching and liveness detection.

The problem

Digital identity verification in low-resource environments faces two core challenges: training accurate facial verification models with constrained localized datasets, and preventing spoofing attacks (printouts, screen replays) without requiring specialized hardware.

Approach

Engineered an end-to-end eKYC platform combining custom biometric models with a scalable SaaS backend. Developed a face-matching pipeline optimized for data-constrained training regimes and implemented active/passive liveness detection using facial geometry and landmark dynamics. The platform runs on a microservices architecture with polyglot storage (relational metadata, caching, and secure blob storage) and exposes an enterprise developer portal for multi-tenant API key management and quota metering.

At a glance

Match + Liveness
Biometrics
Active & passive anti-spoofing
Microservices
Architecture
Decoupled inference & auth services
Polyglot
Storage Model
Relational + cache + encrypted blob
Enterprise SaaS
Deployment
Multi-tenant API & developer portal

Stack

PythonPyTorchComputer VisionMicroservicesDockerFastAPIPolyglot PersistenceAPI Security