Computer-Vision Identity Verification Tool for a Know-Your-Customer (KYC) Compliance Provider

Overview

Built a computer-vision identity verification tool for a KYC compliance provider that checks submitted government ID photos against a live selfie to help confirm an applicant is who they claim to be. The tool checks for common signs of tampering or a photo of a photo, and flags any submission that doesn't clearly pass automated checks for manual review by the client's compliance team rather than approving or denying identity claims on its own. This let the provider handle a growing volume of verification requests without proportionally growing its manual review team. Every automated pass or fail decision is logged with the specific signals behind it, since the client's own regulatory obligations required an explainable audit trail, not a black-box result. A KYC compliance provider needed to handle a growing volume of identity verification requests without proportionally growing its manual review team, while still meeting regulatory requirements for an explainable audit trail. We built a computer-vision tool checking submitted ID photos against a live selfie, flagging tampering signs or photo-of-a-photo attempts, routing anything that doesn't clearly pass to manual review rather than approving or denying identity claims on its own, with every decision logged with its specific signals. We built the detection and logging pipeline around the client's specific regulatory audit-trail requirements from the start, since an unexplainable result would have been a non-starter for their compliance obligations. It was validated against a large batch of known-legitimate and known-fraudulent submissions before deployment. The provider now scales verification volume significantly without a proportional increase in manual review staff, while keeping a fully auditable decision trail that satisfies its regulatory obligations.