eKYC System / Identity verification
From a face to a verified identity.
Liveness checks, face comparison, and document validation in an event-driven verification workflow.
Document and face inputs enter modular preprocessing and validation services.
Architecture illustration based on my project contribution. No customer identity data is shown.
- AWS Rekognition
- AWS Lambda
- ONNX
- GitHub Actions
- AWS CloudFormation
Results & scope
- Images in the training-data workflow
- 20M+
- CV project result
- Supported collection and preprocessing for a face-recognition training dataset during December 2024–August 2025.
Method & conditions
Self-reported in my CV. This is dataset-preparation support, not sole authorship or manual labeling of every image.
- Business-logic unit-test coverage
- 80%
- CV project result
- Reported coverage for the eKYC business-logic layer.
Method & conditions
Self-reported in my CV. Does not describe whole-application or end-to-end test coverage.
How I approached the work
Engineering decisions.
Fit model inference to serverless constraints
- Problem
- Face recognition and document validation needed modular execution within Lambda runtime constraints.
- Decision
- Converted custom recognition models to ONNX and separated image preprocessing, comparison, and validation into Lambda functions.
- Result
- The verification workflow combined custom model inference with Rekognition liveness checks and repeatable infrastructure delivery.
The work.
I contributed to an eKYC platform that combines biometric checks, document validation, and a management portal. My work connected face-recognition models with modular serverless services and the workflows people use to review identity checks.
My contribution
Verification workflows
Integrated AWS Rekognition liveness detection and worked with frontend engineers on the management portal.
Models in serverless runtimes
Supported face-data preparation, optimized models in ONNX, and built Lambda functions for image preprocessing, comparison, and document validation.
Repeatable delivery
Implemented business-logic tests and a GitHub Actions pipeline with CloudFormation infrastructure.