👉 Try the Live Interactive Prototype Here

Catching synthetic fraud and deepfakes silently, before the first API call.

pre-kyc-video1.webm

1. The Landscape & The Problem

In the B2B RegTech space, the standard defense against identity fraud is reactive. Banks and fintechs wait for a user to upload their ID and selfie, and then send that data to a vendor to be verified.

The Enterprise Shortfall: This reactive approach creates massive operational and financial drain for enterprise compliance teams:

1. Financial Waste: Every ID check costs money. If a bot network submits 10,000 fake IDs, the bank pays for 10,000 failed checks.
2. The Deepfake Threat: Generative AI has made synthetic identities indistinguishable from real ones to the naked eye, overwhelming human reviewers and causing severe backlogs.
3. Analyst Burnout: Risk analysts are forced to review thousands of false positives daily through clunky, outdated tabular dashboards, leading to alert fatigue and critical human error.

2. The Hypothesis

What if we could catch the fraudster before they even upload a document? If we build an invisible "Pre-KYC" layer that analyzes device behavior, network anomalies, and interaction patterns the moment a user opens the app, we can block bad actors silently. This saves the client massive API costs and protects their infrastructure from coordinated attacks.

3. The Solution & UX Strategy

I designed the Pre-KYC Risk Engine Dashboard specifically for Enterprise Fraud Analysts. The core UX challenge was translating massive, invisible machine-learning datasets into actionable visual intelligence.

To solve this, I architected a strict dual-layer UI that separates macro-level attacks from micro-level behavior:

• Layer 1: The Fraud Ring Map (Macro Context): Analysts need to see connections, not just isolated events. This view uses node-based data visualization to map relationships between users. If 50 seemingly different accounts share the same hidden device fingerprint or VPN subnet, they are visually clustered into a glowing "Fraud Ring," allowing the analyst to identify coordinated syndicate attacks instantly.
• Layer 2: Live Telemetry Deep-Dive (Micro Context): When an analyst zooms in on a single suspicious session, they enter the Telemetry view. Instead of raw JSON logs, the UI presents a timeline of invisible behavioral flags: e.g., Android Emulator detected, GPS spoofing active, or copy/paste speeds that exceed human capability.
• The "Traffic Light" Heuristic: Analysts have seconds to make a decision. I utilized a strict semantic color system throughout the platform—Crimson (Critical Risk), Amber (Suspicious), and Emerald (Clear)—allowing analysts to process threat levels peripherally.